In recent years, Rydberg atoms have been the subject of intense study,

becoming the testing ground for several quantum mechanical problems.

Thisbook provides a comprehensive description of the physics of Rydberg
atoms, highlighting their remarkable properties by reference to their behavior
in a wide range of physical situations. Beginning with a brief historical
overview, the basic properties, creation and detection of Rydberg atoms are
described. The effects of blackbody radiation are discussed, as are optical
excitation in static electric fields, ionization by pulsed electric fields,
Rydberg spectroscopy in high magnetic fields, and microwave excitation and
ionization. Thecollisionsof Rydberg atomswithneutral atomsand molecules,
charged particles, and other Rydberg atoms are dealt with in detail. The
powerful method of multichannel quantum defect theory is presented, and
usedin the description of autoionizing Rydberg states,interseries interactions
and double Rydberg states. In addition to providing a clear introduction to
the basicproperties of Rydberg atoms, experimental and theoretical research
in this extensive field is reviewed. The books will therefore be valuable to
both graduate students and established researchers in physics and physical
chemistry.

Cambridge Monographs on Atomic, Molecular, and Chemical Physics 3

General editors: A. Dalgarno, P. L. Knight, F. H. Read, R. N. Zare
RYDBERG ATOMS

Cambridge Monographson Atomic, Molecular,and Chemical Physics 1. R.

Schinke Photodissociation Dynamics 2. L. Frommhold Collision-
inducedAbsorption in Gases 3. T. F. Gallagher Rydberg Atoms

RYDBERG ATOMS THOMAS F. GALLAGHER Jesse W. Beams Professor

of Physics Department of Physics, University of Virginia CAMBRIDGE
UNIVERSITY PRESS

CAMBRIDGE UNIVERSITY PRESS Cambridge, New York, Melbourne,

Madrid, Cape Town, Singapore, Sao Paulo Cambridge University Press The
Edinburgh Building, Cambridge CB2 2RU, UK Published in the United States
ofAmerica by Cambridge University Press, New York www.cambridge.org
Information on this title: www.cambridge.org/9780521385312 � Cambridge
University Press 1994 This publication is in copyright. Subject to statutory
exception and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without the written permission of
Cambridge University Press. First published 1994 This digitally printed first
paperback version 2005 A catalogue recordfor this publication is available
from the British Library Library of Congress Cataloguing in Publication data
Gallagher, Thomas F. Rydberg atoms / Thomas F. Gallagher. p. cm. ISBN 0-
521-38531-8 1. Rydberg states. I. Title. QC454.A8S27 1994 539.7�dc20
93-37132 CIP ISBN-13 978-0-521-38531-2 hardback ISBN-10 0-521-
38531-8 hardback ISBN-13 978-0-521-02166-1 paperback ISBN-10 0-521-
02166-9 paperback

Contents Preface xiii 1 Introduction 1 References 9 2 Rydberg atom

wavefunctions 10 References 26 3 Production of Rydberg atoms 27
Excitation to the Rydberg states 27 Electon impact 29 Charge exchange 31
Optical excitation 32 Collisional-optical excitation 35 References 37 4
Oscillator strengths and lifetimes 38 Oscillator strengths 38 Radiative
lifetimes 45 References 49 5 Black body radiation 50 Black body radiation
50 Black body induced transitions 53 Black body energy shifts 55 Initial
verification 57 Temperature dependent measurements 59 Suppression and
enhancement of transition rates 61 Level shifts 64 Experimental
manifestations and uses 65 Far infrared detection 66 References 68 6
Electric fields 70 Hydrogen 70 Field ionization 83 Nonhydrogenic atoms 87
Ionization of nonhydrogenic atoms 95 References 101 vn

viii Contents 7 Pulsed field ionization 103 Hydrogen 104 Nonhydrogenic

atoms 105 Spin orbit effects 115 References 119 8 Photoexcitation in electric
fields 120 Hydrogenic spectra 120 Nonhydrogenic spectra 135 References
141 9 Magnetic fields 143 Diamagnetism 143 Quasi Landau resonances 147
Quasi classical orbits 153 References 160 10 Microwave excitation and
ionization 162 Microwave ionization of nonhydrogenic atoms 164
Microwave multiphoton transitions 168 Hydrogen 182 Ionization by
circularly polarized fields 190 References 193 11 Collisions with neutral
atoms and molecules 195 A physical picture 196 Theory 198 Experimental
methods 205 Collisional angular momentum mixing 208 The effects of
electric fields on � mixing 212 � mixing by molecules 213 Fine structure
changing collisions 215 n changing collisions by rare gases 216 n changing
by alkali atoms 220 n changing by molecules 221 Electron attachment 230
Associative ionization 239 Penning ionization 243 Ion-perturber collisions
243 Fast collisons 245 References 247

Contents ix 12 Spectral line shifts and broadenings 250 Theory 250

Experimental methods 254 Measurements of shifts and broadening 255
References 267 13 Charged particle collisions 269 State changing collisions
with ions 269 Na nd �� ra�transitions 270 Depopulation of the Na ns and
np states 272 Theoretical descriptions 275 Electron loss 276 Charge
exchange 279 Electron collisions 285 References 288 14 Resonant Rydberg-
Rydberg collisions 290 Two state theory 294 Box interaction strength
approximation 295 Exact resonance approximation 296 Numerical
calculations 298 Intracollisional interference 301 Calculation of cross
sections 301 Experimental approach 302 n scaling laws 303 Orientation
dependence of v and E and intracollisional interference 306 Velocity
dependence of the collisional resonances 307 Transform limited collisions
312 References 313 15 Radiative collisions 314 Initial experimental study of
radiative collisions 317 Theoretical description 321 References 339 16
Spectroscopy of alkali Rydberg states 340 Optical measurements 341
Radiofrequency resonance 343 Core polarization 348 Fine structure intervals
352 Quantum beats and level crossing 355 References 363

x Contents 17 Rf spectroscopy of alkaline earth atoms 365 Theoretical

description of series perturbation and nonadiabatic 18 19 20 21 22 effects in
core polarization Experimental methods Low angular momentum states and
series perturbations Non adiabatic core polarization References Bound
HeRydberg states Theoretical description Experimental approaches Quantum
defects and fine structure References Autoionizing Rydberg states Basic
notions of autoionizing Rydberg states Experimental methods Experimental
observations of autoionization rates Electron spectroscopy References
Quantum defect theory Quantum defect theory (QDT) Geometrical
interpretation of the quantum defect surface Normalization Energy constraints
Alternative R matrix form of QDT The role of QDT References Optical
spectra of autoionizing Rydberg states The shake up satellites Interacting
autoionizing series Comparison between experimental spectra and R matrix
calculations References Interseries interaction in bound states Perturbed
Rydberg series Properties of perturbed states Continuity of photoexcitation
across ionization limits Forced autoionization References 365 373 375 376
380 381 381 384 391 393 395 395 399 408 411 413 415 415 421 422 423
424 427 428 429 434 440 448 451 453 453 457 460 461 464

Contents xi 23 Double Rydberg states 466 Doubly excited states of He 467

Theoretical descriptions 469 Laser excitation 480 Experimental observations
of electron correlation 482 References 491 Index 493

Preface My intent in writing this book is to present a unified description of

the many properties of Rydberg atoms. It is intended for graduate students
and research workers interested in the properties of Rydberg states of atoms
or molecules. In many ways it is similar to the excellent volume Rydberg
Statesof Atoms and Molecules edited by R. F. Stebbings and F. B. Dunning
just over adecade ago. It differs, however, in covering more topics and in
being written by one author. I have attempted to focus on the essential
physical ideas. Consequently the theoretical developments are not
particularly formal, nor is there much emphasis on the experimental details.
The constraints imposed by the size of the book and myenergy have forced
me to limit the topics covered in thisbook to those of general interest and
those about which I already knew something. Consequently, several
important topics which might well have been included by another author are
not included in the present volume. Two examples are molecular Rydberg
states and cavity quantum electrodynamics. Finally, itis agreat pleasure to
acknowledge the fact that thisbook would never have been written without
the efforts of many people. First I would like to acknowledge the help of my
colleagues in the Molecular Physics Laboratory of SRI International
(originally Stanford Research Institute). They had the confidence that our
initial experiments would develop into a productive research program, and
they completed my education as a physicist. My colleagues at the University
of Virginia have continued to provide both a critical audience and the
encouragement necessary to undertake the writing of this book. My
collaborators have contributed substantially to my understanding of Rydberg
atoms, and it is a pleasure to acknowledge the contributions of L. A.
Bloomfield, W. E. Cooke, S. A. Edelstein, F. Gounand, R. M. Hill, R.
Kachru, R. R. Jones, D. J. Larson, D. C. Lorents, L. Noordam, P. Pillet, K. A.
Safinya, W. Sandner, and R. C. Stoneman. In addition I have been the
beneficiary of the insights of my students, post doctoral fellows, and visitors.
Without all of their contributions this book could not have been written. I am
indebted to Tammie Shifflett, Bessie Truzy, Warrick Liu, and Sibyl Hale for
their careful typing of the manuscript, and it is apleasure to acknowledge the
encouragement ofJamesDeeny, Rufus Neal, andPhilip Meylerofthe
Cambridge xni

xiv Preface University Press. Finally, the gentle prodding of my mother,

Margaret Gallagher, and the patience of my wife, Betty, played important
roles in the completion of this book. Charlottesville, Virginia T. F. Gallagher
November, 1993

1 Introduction Rydberg atoms, atoms in states of high principal quantum

number, n, are atoms with exaggerated properties. While they have only been
studied intensely since the nineteen seventies, they have played a role in
atomic physics since the beginning of quantitative atomic spectroscopy. Their
role in the early days of atomic spectroscopy is described by White.1 The
first appearance of Rydberg atoms is in the Balmer series of atomic H.
Balmer'sformula, from 1885, for thewavelengthsofthevisible series
ofatomicH, is given by1 where b = 3645.6A. We nowrecognize Eq. (1.1) as
givingthewavelengths of the Balmer series of transitions from the n = 2states
to higher lying levels. While the H atom was the first atom to be understood
in a quantitative way, other atoms played a crucial role in unravelling the
mysteries of atomic spectroscopy. For example, Liveing and Dewar2made
theimportant observation that the observed spectral lines of Na could be
grouped into different series. Specifically they observed the Na ns-3p and
nd-3p emissions from an arc. Up to the 9s and8d states they observed
doublets due to thefinestructure splitting of the3p state, but for the 10s and 9d
states they were unable to resolve the 3p doublet. The ns-3p doublets
appeared as sharp lines while the nd-3p doublets appeared as diffuse lines.
Most important, they recognized that these sharp and diffuse doublets were
members of two series of related lines, although they were unsuccessful in
their attempts to relate the wavelengths of the observed transitions. Knowing,
as we do today, that the Na d states are almost degenerate with the higher
angular momentum states of the same n, we are not surprised that the series of
transitions to the d Rydberg series is diffuse. They are much more likely to
suffer pressure broadening. In contrast to Na, Liveing and Dewar noted that
in the K spectrum the s and d serieswerenotparticularly different, both being
"more or less diffuse".2 We now realize that the difference between Na and
K is due to the fact that both the s and the d Rydberg states of K are
energetically well removed from the high angular momentum states. A crucial
advance was made byHartley, whoinhis studies of thespectra ofMg, Zn, and
Cd, first realized the significance of the frequency of a transition as opposed
to the experimentally measured wavelength.1'3 Hartley observed that 1

2 Rydberg atoms the splittings of series of multiplets always had the same
wavenumber splitting, irrespective of the wavelengths of the transitions. The
wavenumber, v, is the inverse of the wavelength in vacuum. The importance
of thisrealization is hard to overstate and isimmediately apparent if
werewrite Balmer's formula in termsof the wavenumber of the observed
lines instead of the wavelength, It is evident that it simply reflects the energy
difference between the n = 2 and higher lying states. Following the precedent
of Liveing and Dewar, Rydberg began to classify the spectra of other atoms,
notably alkali atoms, into sharp, principal, and diffuse series of lines.4 Each
series of lines has a common lower level and a series of ns, np, or nd levels,
respectively, as the upper levels of the sharp, principal, and diffuse series.
He realized that the wavenumbers of the series members were related and
that the wavenumbers of the observed lines could be expressed as1'4 VQ =
Voo* - � ( L 3 ) Ry where the constants v^, Voop, and Vood and (5S,(5p, and
dd are the series limits and quantum defects of the sharp, principal, and
diffuse series. The constant Ry is a universal constant, which can be used to
describe the wavenumbers of the transitions, not only for different series of
the same atom, but for different atoms as well. This constant, Ry, is now
known as the Rydberg constant. Defining the wavenumber as the inverse of
the wavelength in air, Rydberg4 assigned Ry the value 109721.6 cm"1.
Recognition ofthegenerality ofthe Rydberg constant is one of Rydberg's two
major accomplishments. The other was to show that all the spectral lines
from an atom are related. Specifically, he realized that VooS and Vood were
the same to within experimental error. For example, for the series originating
in the Na 3p3/2 and 3p1/2 levels he assigned the values 24485.9 and
24500.5 cm"1 for Voos, and 24481.8 and 24496.4 cm"1 for Vood. In
addition, heobserved that Voop � VooSwas equal tothewavenumber of
thefirstmember oftheprincipalseries. He reasoned that it should be possible
to see other series which did not have the first term of the principal series in
common, but some other term instead. This notion led him to a general
expression for the wavenumbers of lines connecting different series. The
wavenumbers of lines connecting the s and p series, for example, are given
by4

Introduction & Fig. 1.1 Energy level diagram for Na showing the sharp 3p-
ns, principal 3s-np, and diffuse 3p-nd series. � v= R y R y (m-ds)2 (n-dp)2
In Eq. (1.4) the + sign and constant n describe a sharp series of s states and
the minus sign and a constant m describe a principal series of p states. If we
consider the special case ds = dp =0 and m =2 we arrive at Balmer's formula
for the H transitions from n=2. From the early spectroscopic work it is
possible to construct an energy level diagram for Na, as shown in Fig. 1.1.
From this figure it is apparent that the difference in the principal and sharp
series limits is the wavenumber of the 3s-3p transition. It is also apparent
that the Rydberg states, states of high principal quantum number n, lie close
to the series limit. The physical significance of high ndid not become clear
until Bohrproposed his model of the H atom in 1913.l In spite of
itsshortcomings, theBohr atom contains the properties of Rydberg atoms
which make them interesting. The basis of the Bohr atomis anelectron
movingclassicallyinacircular orbit around anioniccore. To the currently
accepted ideas of classical physics Bohr added two notions; that the angular
momentum was quantized in integral units of ft, Planck's constant divided by
2jr, and that the electron did not radiate continuously in a classical manner,
but gave off radiation only in making transitions between states of well
defined energies. The first was an assumption, and the second was forced
upon him by the experimental observations.

4 Rydberg atoms The Bohr theory can be summarized as follows. An electron

of charge �e and mass m in acircular orbit of radius r about an infinitely
heavy positive charge of Ze obeys Newton's law for uniform circular
motion5 .,2 ^ = ^ - , (1.5) r r where k = l/4jie0, e0 being the permittivity of
free space. Adding Bohr's requirement of the quantization of angular
momentum yields mvr = nh. (1-6) Combining Eqs. (1.5) and (1.6) leads
immediately to an expression for the radius r of the orbit (1.7) Ze2mk In
other words, the size of the orbit increases as the square of the principal
quantum number, and states of high n have very large orbits. The energy, W,
of a state is obtained by adding its kinetic and potential energies, W = � - ^ -
= , ' � (1.8) 2 r 2n2h2 K ' The energies are negative, i.e. the electron is
bound to theproton, and the binding energy decreases as 1/n2.The allowed
transition frequencies are the differences in the energies. Explicitly \ - \ \ (1-
9) Comparing this expression to the general form of Rydberg's formula we
can see that Ry = Historically, the most important aspect of the Bohr atom
was that it related the spectroscopic Rydberg constant to the mass and charge
of the electron. From our present point of view, the Bohr atom physically
defines Rydberg atoms and shows us why they are interesting. In high n, or
Rydberg, states the valence electrons have binding energies which decrease
as 1/n2 and orbital radii which increase as n2. In other words, in a Rydberg
atom the valence electron is in a large, loosely bound orbit. Although
Rydberg atoms of principal quantum number well over 100 have been
studied inthe laboratory, itis instructive to consider a relatively low Rydberg
state, n = 10, and compare it to aground state atom. The ground stateof H
isbound by 1 Ry, 13.6eV, and has an orbital radius of la0 and ageometric
cross section of a02' In contrast, the n = 10 state has a binding energy of 0.01
Ry, an

Introduction n = 10 Fig. 1.2 Bohr orbitsof n = 1 and n = 10. The solid dark
circle at the center is then � \ orbit of radius la0. The n = 10orbit has radius
100a0. la0 =0.53 A. orbital radius of 100a0, and a geometric cross section of
104a02. Fig. 1.2 is a drawing of the Bohr orbits for n = 1 and n = 10,which
demonstrates graphically the difference between a Rydberg atom and a
ground state atom. The binding energy of the n = 10 atom is comparable to
thermal energies, and the geometric cross section is orders of magnitude
larger than gas kinetic cross sections. Once the Bohr theory was formulated it
was apparent that Rydberg atoms should have bizarre properties. Why then
were Rydberg atoms not studied more extensively until the nineteen
seventies? There are two reasons. First, the most pressing problem at the
time wasnot the properties of atoms near the series limit, but the development
of the quantum theory. Second, there was simply no means of making Rydberg
atomsefficiently enough to studytheminthe detailthat isnow possible.
Nonetheless, the most sophisticated tool available at the time, high resolution
absorption spectroscopy, was used in the first experiments to demonstrate the
bizarre properties of Rydberg atoms. By examining the fine details of the
spectra, the splittings, shifts, and broadenings, it waspossible to learn agreat
deal about Rydberg states. One of thefirst properties to be explored was the
sensitivity of Rydberg atoms to external electric fields, the Stark effect.
While ground state atoms are nearly immune to electric fields, relatively
modest electric fields not only perturb the Rydberg energy levels, but even
ionize Rydberg atoms, as was shown in early

Rydberg atoms -0.02 Fig. 1.3 CombinedCoulomb-Starkpotential alongthe

zaxisfor afieldof5.14 kV/cmin the z direction. studies of the Balmer series in
electricfieldsof up to 106V/cm.6'7The Balmer lines exhibit a splitting which
is approximately linear in the electric field, and the components actually
disappear at well defined values of the electric field due to field ionization
of the Rydberg state. While the Bohr atom is of no help in understanding the
splittings of the Balmer lines, using it we can calculate the field at which a
state is ionized by an electric field. Consider aH atom withits nucleus at the
origin in the presence of an electric field in the z direction. The potential
experienced by an electron moving along the z axis is given by V=- � + Ez
(1.11) and is shown in Fig. 1.3. Only electrons with energies lower than the
local maximum of the potential, at z = � 1000a0, are classically bound. In
the real three dimensional potential the local maximum at z = -1000tf0 is a
saddle point, and electrons with energies above the saddle point of the
potential are ionized by the field. The potential at the saddle point is given by
Vs=-2VkE. (1.12) If this energy is equated to the energy of the Rydberg state,
-Ry/n2, we find the classical field for ionization of a state of principal
quantum number n This formula is usually encountered in atomic units, in
which it reads Ec=lll6n\ (1.14) In the next chapter weshallintroduce
atomicunits. Although we have derived Eq. (1.13) with no regard for
tunneling or the shifts of the energy levels in an electric

Introduction 7 field,it is quite a useful expression, as it is, in a practical

sense, nearly exact for nonhydrogenic atoms. The size of a Rydberg atom is
large, the geometric cross section scaling as n4, and, not surprisingly, early
experiments were focused on this aspect of Rydberg atoms. A classic
experiment wasundertaken by Amaldi and Segre to observe the energy shifts
of the high lyingmembers of the Rydberg series of K in the presence of high
pressures of rare gas atoms.8 Specifically, they observed the shifts of the 4s-
np absorption lines due to the added rare gas. The expectation at the timewas
that a red shift of the lines would be observed due to the fact that the space
between the Rydberg electron andthe K+ ionisfilledwith apolarizable
dielectric, the rare gas atoms, instead of vacuum. While the pressure shifts
for Ar and He were to the red, the pressure shift for Ne was to the blue, an
observation completely incompatible with the dielectric model. Shortly after
the experiment Fermi explained the pressure shifts asarising predominantly
from the short range scattering of the Rydberg electron from the rare gas
atom, not from the anticipated dielectric effect.9 This experiment was
perhaps the first of many in which the result was diametrically opposed to
expectations. Due to their large size, Rydberg atoms also exhibit large
diamagnetic energy shifts. Since the Zeeman and Paschen-Back effects are
proportional to the angular momentum, they are not particularly striking in the
optically accessible low � Rydberg states. The diamagnetic shifts on the
other hand depend on the area of the orbit and thus scale as n4. The
diamagnetic shifts are very difficult to see in low n states, but in high n states
they are quite evident. The diamagnetic shifts were first observed by Jenkins
and Segre in the absorption spectrum of K taken in the 27 kGfieldof the
Berkeley cyclotron.10 As n increases thefieldfirst mixesstates of different
�,then mixes states of different n. Only within the last20
yearshaveexperiments progressed substantially beyond thesefirstexperiments
of Jenkins and Segre. Ironically, the study of Rydberg atoms in
magneticfields,one of the first topics studied, is still a topic of active
research 50years later. The interest in Rydberg atoms waned until it became
apparent that they might play an important role in real physical systems. In
particular it became apparent that they were important astrophysically.11 The
radio recombination lines detected by radio astronomers are the emissions
from transitions between Rydberg states subsequent to radiative
recombination. A particularly favorable frequency regime in which to search
for radio signals is 2.4 GHz, which correspondsto the n = 109 ton = 108
transition. Whileitis not surprising that Rydberg atoms might be found
ininterstellar space,where the density isverylow,they also play an important
role in astrophysical and laboratory plasmas.1213 Radiative recombination
of low energy electrons and ions results in Rydberg atoms, and dielectronic
recombination of higher energy electrons and ions often proceedsby the
capture of the electrons into autoionizing Rydberg states and the subsequent
radiative decay to bound Rydberg states. Furthermore, the microscopic
electric fields in aplasma cut off the Rydberg series at ornear the
classicalionization limit,

8 Rydberg atoms as described by Eq. (1.4). Precisely how the Rydberg

series is terminated is a matter of some importance, for it determines the
thermodynamic properties of the plasma.12 The single event which probably
contributed the most to the resurgence of interest in Rydberg atomswas the
development ofthe tunable dyelaser.14'15 With a tunable dye laser and very
simple apparatus it became possible to excite large numbers of atomsto
asingle,welldefined Rydberg state. This ability allowed one to study a wide
range of properties in astonishing detail. Manyproperties familiar from
lowlyingexcited states, such astheradiative lifetimes, energylevel spacings,
and collision cross sections were measured systematically, and in many, but
not all, cases the properties followed the expected extrapolations from the
low lying states. Perhaps more interesting were the investigations of the
properties which are in a practical sense peculiar to Rydberg atoms. A good
example of such properties is the effect of an electricfield.While the gross
effects of electric fields on Rydberg atoms had been observed long before in
optical spectra, and worked out theoretically for some cases, they had not
been understood in a systematic way,mainlybecause there appeared
tobenopressingreason todoso. Not onlydo electricfieldsproduce dramatic
effects on Rydberg atoms, but they also afford us novel techniques for the
detection and manipulation of Rydberg atoms. To exploit fully these
techniques has required a systematic study of the static and dynamic effects
of electric fields. The most fascinating opportunities presented by Rydberg
atoms are those which take advantage of their exaggerated properties to carry
out experiments that would be impossible in other systems. One example of
this is the interaction of atoms with radiation fields. In the low intensity limit,
Rydberg atoms provide an ideal system with which to test the interaction of
an atom with the vacuum, particularly with the structured vacuum produced
bya resonant cavity. Outstanding examples are the one atom maser16 and the
two photon maser.17 Rydberg atoms have two properties which make them
ideal for such investigations. First, the characteristic frequencies of Rydberg
atom transitions are low, and the wavelengths correspondingly long, making
the construction of resonant or near resonant cavities a straightforward
matter. Second, the dipole moments of Rydberg atoms are so large that they
have usable radiative decay rates in spiteof the low transition frequencies. At
the other intensity extreme, advances in laser technology have made atoms in
strong radiation fields, fields comparable in magnitude to the coulomb field,
a topic of both practical and fundamental interest.18"20 While it is apparent
that any practical applications will involve intense lasers and ground state
atoms and molecules, a Rydberg atom in a microwave field is an ideal
system in which to carry out quantitative measurements of the properties of
atoms in strong radiation fields.21 All the properties we have discussed thus
far are those of atoms with one Rydberg electron. Double Rydberg atoms,
atoms in which there are two highly excited electrons, exhibit pronounced
correlation between the motions of thetwo

References 9 electrons and have properties which differ in a qualitative way

from those of normal Rydberg atoms.22 While some experiments have been
done, they have onlyscratched the surface, andmuchisyetto belearned about
theseexotic atoms. References 1. H. E. White, Introduction to Atomic Spectra
(McGraw-Hill, New York, 1934). 2. G. D. Liveing and J. Dewar, Proc. Roy.
Soc. Lond. 29,398 (1879). 3. W. N. Hartley, /. Chem. Soc. 43,390(1883) 4. J.
R. Rydberg, Phil. Mag. 5th Ser. 29, 331 (1890) 5. H. Semat and J. R.
Albright, Introduction to Atomic and Nuclear Physics (Holt, Rinehart, and
Winston, New York, 1972). 6. H. A. Bethe and E. A. Salpeter, Quantum
Mechanics of One and Two Electron Atoms (Academic Press, New York,
1957). 7. H. Rausch v. Traubenberg, Z. Phys. 54307 (1929). 8. E. Amaldi
and E. Segre, Nuovo Cimento 11, 145 (1934). 9. E. Fermi, Nuovo Cimento
11, 157 (1934). 10. F. A. Jenkins and E. Segre, Phys.Rev. 55, 59 (1939). 11.
A. Dalgarno, in Rydberg States of Atoms and Molecules, eds. R.F. Stebbings
and F.B. Dunning (Cambridge University Press, Cambridge, 1983). 12. D. G.
Hummer and D. Mihalis, Astrophys. J. 331, 794,(1988). 13. V. L. Jacobs, J.
Davis, and P. C. Kepple, Phys.Rev. Lett.37, 1390,(1976). 14. P. P. Sorokin
and J. R. Lankard, IBM3. Res. Dev. 10, 162,(1966). 15. T. W. Hansch, Appl.
Opt. 11, 895,(1972). 16. D. Meschede, H. Walther, and G. Miiller, Phys.
Rev. Lett. 54, 551, (1985). 17. M. Brune, J. M. Raimond, P. Goy, L.
Davidovitch, and S. Haroche, Phys.Rev. Lett.59,1899, (1987). 18. A.
L'Huillier, L.A. Lompre, G. Mainfray, and C. Manus, Phys. Rev. Lett.48,
1814,(1982). 19. T. S. Luk, H. Pommer, K. Boyer, M. Shahidi, H. Egger, and
C. K. Rhodes, Phys. Rev. Lett. 51, 110,(1983). 20. R. R. Freeman, P. H.
Bucksbaum, H. Milchberg, S. Darack, D. Schumacher, and M. E. Geusic,
Phys. Rev. Lett. 59, 1092, (1987). 21. J. E. Bayfield and P. M. Koch,
Phys.Rev. Lett. 33, 258, (1974). 22. U. Fano, Rep. Prog. Phys.46, 97,(1983).

2 Rydberg atom wavefunctions To calculate the properties of Rydberg atoms

we need wavefunctions to describe them. In thischapter we usethe methods of
quantum defect theory to develop the wavefunctions for an electron in a
coulomb potential. This approach may be applied to any single valence
electron Rydberg atom, including the H atom as a special case. From a
practical point ofviewit isanimportant special case, because the analytic
properties of the hydrogenic wavefunctions are well known and are useful in
obtaining analytic solutions to a variety of problems. We also present
numerical methods for generating accurate coulomb wavefunctions for use in
later problems. Finally, we showhow scaling lawsof many properties of
Rydberg atoms are easily obtained from the wavefunctions. If we consider
Rydberg states of H and Na, as shown in Fig. 2.1, they are essentially
similar. The only difference between the two is that the Na+ core, while
having charge 1, has a finite size due to its being made up of a nucleus of
charge +11 and ten electrons. When the Rydberg electron is far from the Na+
core it isonly sensitive to the net charge. Since the Rydberg electron spends
most (a) (b) Fig. 2.1 Rydberg atoms of (a) H and (b) Na. In H the electron
orbits around the point charge of the proton. In Na it orbits around the +11
nuclear charge and ten inner shell electrons. In high �states Na behaves
identically to H, but in low�states the Na electron penetrates and polarizes
the inner shell electrons of the Na+ core. 10

Rydbergatom wavefunctions 11 of itstime near itsclassical outer turning

point, where the differences between Na and H are minimal, we expect the
properties of all Rydberg atoms to be similar. On the other hand, when the
Rydberg electron comes near the Na+ core the precise charge distribution of
the Na+ core does play a role. In particular, the Rydberg electron can both
polarize and penetrate the Na+ core, and both polarization and penetration
change the wavefunctions and energies of a Na Rydberg state from their
hydrogenic counterparts. If we consider Na and H atoms in states of high
orbital angular momentum, for example in circular Bohr orbits, the
differences are negligible, for the Na electron never comesclose enough to
the Na+ core to detect that it is anything other than a point charge. On the
other hand, ifthe electron is inahighlyellipticlow�orbit, on each orbit the
Rydberg electron comes close to the core, where there is a significant
difference between Na and H. In fact, there are very interesting differences
between the low �and high �states of Na, the most obvious being the
depression of the energies of the low � states below the H levels, due to
core polarization and penetration. When the Na Rydberg electron penetrates
the cloud of the inner ten electrons, it is exposed to the unshielded +11
nuclear charge, and its binding energy is increased, or, equivalently, its total
energy decreased. Similarly, polarization of the Na+ core produces a
decrease in the energy of a Rydberg state. In Fig. 2.2 we show the energy
levels of H and Na. As expected, the H and high �Na levels are degenerate
on the scale of Fig. 2.2, but the low �states of Na are depressed in energy.1
Their energies are given by (2.1) (� - 6eY where d� is an empirically
observed quantum defect for the series of orbital angular momentum �. The
depression of the energies of the low � states of Na below the H values is
perhaps the most obvious difference between Na and H. However the
alteration in the wavefunction required to produce these energy differences
leads to many other differences aswell. In the calculation of atomicproperties
itis convenient tointroduce atomicunits which are defined so that all the
relevant parameters for the ground state of H have magnitude one. The atomic
units most useful for our purposes are given in Table 2.1.2An extensive list is
given by Bethe and Salpeter.2 Throughout the book atomic units willbe used
for all calculations, with conversions to other units to facilitate comparisons
to experiment. In developing Rydberg atom wavefunctions we begin with the
Schroedinger equation for the H atom, which, in atomic units, may be written
as V2 1\ * '"� ( 2 . 2 ) where r is the distance of the electron from the
proton, assumed to be infinitely massive, and W is its energy. Unless stated
otherwise, we shall consider only

12 Rydberg atoms SODIUM HYDROGEN s P d f 9 __ _ 7_ 8 7_ 7 6 8 7 6 5

8 7 6 5 8 7 6 5 �_ 10 IFig. 2.2 Energy level diagram for H and Na (from ref.
1). Table 2.1. Atomic units. Quantity Atomic unit Definition Mass Electron
mass 9.1 x 10~28 g Charge Electron charge e 1.6 x 10~19C Energy Twice
the ionization potential of hydrogen 27.2 ey Length Radius of a0,the first
Bohr orbit 0.529A Velocity Velocity of thefirstBohr orbit 2.19 x 108 cm/s
Electric field Field at thefirstBohr orbit 5.14 x 109V/cm
Rydberg atom wavefunctions 13 x Fig. 2.3 Geometry of the atom. neutral
atoms in which an electron far from the ionic core sees a charge of +1. In
spherical coordinates J 4v++A(sin,A)+ 4 , (2.3) dr2 rdr r2 sin 6 dO\ dOj r2
sin2 0 dcp2 where 0 and 0 are the polar angle relative to the quantization
axis and the azimuthal angle, respectively, as shown in Fig. 2.3. If we assume
that Eq. (2.2) is separable and write rp as a product of radial and angular
functions, i.e. V = Y(6,(f>) R(r), Eq. (2.2) becomes 2(w+l)R]Y+ dr2 r dr =
0. (2.4) [r2 sin 0 d 0 \ dO) r2 sin2 0 d2 <p Dividing by RYlr2 yields the
separated form r2\d2R 2dR ^/ l \ o l 1 [ 1 d ( . ndY\ 1 dY] _ � �T + +
21W+�\R 4� sin0� +�^ T- =0. ^[ar2 r a r \ r/ J Y[sin^a(9\ d0/ sin20d20J
(2.5) The two terms, which depend only on R and Y, respectively, are
independent of each other and therefore must be separately equal to �A,
where X is a constant. If we further assume that we can write the angular
function as Y(0,0) = O(0)<E>(0), we may write the angular equation as
sin0d0\ dO] shr 0 d <p The solution to Eq. (2.6) is the normalized spherical
harmonic,2

14 Rydberg atoms Table 2.2. Normalized spherical harmonicsfor s,p, and

dstates.a Y10 = Y1�1 = Y20 = Y2�i = Y2�2 = 1 ~\8JrSm J5(fcos2*-f) �
/� sin 0 cos 6 e�l<t> \ oJT "(fromref. 2) - ^ | ^ ^ *Y" (cos ey�*, (2.7)
where PT(x) is the unnormalized associated Legendre polynomial, ( is zero
or a positive integer, m takes integral values from -� to �, andA = �(� +
1). The spherical harmonics are normalized, sin 0 d0 f" d0Y?
m(0,0)y�m(0,0) = 1. (2.8) In Table 2.2 we have listed the first few
spherical harmonics, for the s, p, and d states. It is worth noting that some
authors introduce a factor of [- l]m in defining the associated Legendre
polynomials, producing a corresponding difference in the spherical
harmonics.2 There are �-ra nodes in the 6 coordinate, and none in the 0
coordinate. The angular equation requires X = �(� + 1)where�is
apositive integer. Using this value for A the radial equation of Eq. (2.5) can
be written as d2R 2 dR [\ 2 �(� + 1)1n A ��- + + \2W + v o } \R = 0.
(2.9) dr2 r dr [ r r2 \ ' We now convert the radial equation to the standard
form of the coulomb problem by introducing p{f), which is defined by3 R{r)
= p(r)/r. With this substitution the radial equation of Eq. (2.9) may be written
as
Rydbergatom wavefunctions 15 CD LLJ Fig. 2.4 Coulomb and centrifugal
potentials (�) which are summed to give the effective radial potential (�)
in which an electron of angular momentum � moves. The terms in the square
brackets of Eq. (2.10) have the significance shown inFig. 2.4: 2/r is twice the
coulomb potential of the electron, [�(�+l)]/r2is twice the centrifugal
potential, and 2Wis twice the total energy. The sum in the brackets is 2T,
where Tis the radial kinetic energy of the electron. In the classically allowed
region T > 0, and Eq. (2.10) has two independent oscillatory solutions much
like sines and cosines. On the other hand, in the classically forbidden region
T < 0, and the solutions to Eq. (2.10) are increasing and decreasing
exponential functions. Since Eq. (2.10) is a second order equation it has two
independent solutions. The two solutions of most use in solving our problem
are4 and (2.11) The / and g functions are commonly termed the regular and
irregular coulomb functions.4'5 In the classically allowed region the / and g
functions are real

16 Rydberg atoms oscillatory functions with a phase difference of 90

degrees. In the classically forbidden regions they are increasing or
decreasing exponentials. To find the solutions of the radial equation the
forms of the/and gfunctions as r �> 0 and r �> oo are useful. For W > 0 and
W < 0, as r-� 0 /(W,�,r) <*/+1 (2.12a) and g(WAr) <* r~e. (2.12b) From
Fig. 2.4 it is apparent that the small rbehavior is not very energy dependent,
because T is very large, except for very high angular momentum. For W < 0
we introduce the effective quantum number v, defined by W = �1/2 v2. For
r�> oo the / and g functions may be expressed in terms of increasing and
decreasing exponential functions w(v,�,r) and v(v,�,r). Asr��oo? ^�>
oo andv^ 0, and/and g are given by4 /-> u(v,�,r) sin JTV - v(v,�,r)e^,
(2.13a) g^ -u(y,e,r) cos jrv + v(v,�,r)e>*(v+1/2>. (2.13b) For W> 0 and r
�� oo the/and g functions are given by the oscillatory functions /-> (2/kjt)m
sin f"fcr- jr�/2 +- �n{Her) + aJ , (2.14a) g-> -(2/kjt)1/2 cos A:r - jr�/2 +
- �n (2fcr) + aA , (2.14b) where k = (2W)1/2 and a^ is the coulomb phase,
a� = arg F(� + 1- i/A:). To find bound state solutions, W < 0, for the H
atom we apply the r � 0 and r = ooboundary conditions. Specifically we
require that xp be finite as r�> 0 and that xp�>0 as r�> oo. We can see
from Eqs. (2.12) that only the/functions are allowed due to the r = 0 boundary
condition. As r-� oo we require that xp->0, and, as indicated by the
asymptotic form of the /function, this requirement is equivalent to requiring
that sin jtv be zero or that v be an integer. Combining the angular function of
Eq. (2.7) with the / radial function yields the bound H wavefunction 7 ^ ,
(2.15) where the radial function is assumed to be normalized. Normalization
proceedures will be discussed shortly. Since W = �1/2v2 and v must be
apositive integer /i, we recover the familiar expression for the allowed
energies, W� �. (2.16) If the radial equation is solved by using the more
conventional power series solution,2 the two mathematically allowed
solutions for R(r) have leading terms

Rydbergatom wavefunctions 17 -1.0 4 6 8 10 12 14 Fig. 2.5 Schematic

coulomb potential of H ( ) and VNa(r), the effective one-electron potential
ofNa, ( ) which is different from thecoulomb potentialonly atsmall r (from
ref. 1). proportional to / and r � 1 .Ther � l solution, the irregular coulomb
function, is discarded since it diverges at the origin and does not satisfy the
r=0 boundary condition. The r�> o� boundary condition imposesthe
requirement that the series terminate, which in turn requires that W = -l/2n2,
where n is an integer. The power seriessolution alsoshows that,
theradialfunction Rne(r) hasn � � nodes. If we choose solutions for Rn^(r)
oc / at r ~ 0, the outer lobe of the wavefunction changes sign as either n or
�is increased by one. The above treatment of H, based on coulombwaves,
maybe easily extended to generate wavefunctions for single valence electron
atoms with spherical ionic cores. This approach, quantum defect theory,5
enables us to generate wave functions which are good when the valence
electron is outside the ionic core. We first recall the difference between a Na
atom and a H atom, the replacement of the point charge of the proton by the
finite sized ionic core of the Na atom. Far from the Na+ core the potential
experienced by the valence electron is indistinguishable from the coulomb
potential due to aproton. However, at small orbital radii, where the outer
electron canpenetrate the ten electron cloud of the Na+ core, the Na potential
is deeper than the coulomb potential. Since the Na+ core is spherically
symmetric, and we assume it to be frozen in place, the effective potential,
VNa seen by the valence electron is spherically symmetric and depends
onlyon r. When the coulomb - Vrpotential ofEq. (2.2)isreplaced by
yNa(r),Eq. (2.2) isstill separable, and the angular equation, Eq. (2.6), and its
solutions, Eq. (2.7), are unchanged. Consequently the Na wavefunction
analogous to the H wavefunction of Eq. (2.15) differs only in its radial
function. In Fig. 2.5 we show in a schematic way the coulomb �llr potential
and the effective potential VNa(r) seen bythe Rydberg electron intheNa atom.
The effect

18 Rydberg atoms Hydrogen Fig. 2.6 Wavefunctions of H and alkali atoms.

The lower potential of the alkali introduces the phase shift r, as shown,
leading to the depression of the energies of alkali, low �states relative to
those of H (from ref. 1). of the lower potential at small r seen by thevalence
electron inNais to increase its kineticenergy and decrease the wavelength of
the radial oscillations relative to H. InNa allthe nodesofthe
radialwavefunction arepulled closertotheorigin thanin H, asshown by Fig.
2.6. For r > r0, the radius of the Na+ core, the Na potential is equal tothe
coulomb potential. Thus inthis region the wavefunction is simply shifted
inphase from the Hwavefunction. The magnitude of theradial phase shift isthe
difference in momentum of an electron of energy Win Na and H integrated
from r = 0 to r0. For an s electron, the phase shift rrelative to H can
bewritten as T= )[w-VNa(r)]1/z-\W+-\ V2dr. (2.17) Jo I L r\ } The terms
inthe square brackets are the kinetic energies of the electron inNa and H. If
we replace VNa(r) by �\lr � Vd(r), where Vd is the difference between the
Na and H potentials, and take advantage of the fact that for r < r0 the kinetic
energy islarge so that llr � Wand Vd(r), we can expand Eq. (2.17). Keeping
the first order term in the expansion leads to \l/2 Thus r, the radial phase
shift, is independent of the energy W as long as \W\ � l/r0. A phase shift of
rinthe radial wave from H implies that for r> r0, the pure /wave of the H
wavefunction of Eq. (2.15) is replaced by /(W,�,r)cos r - g(W,�,r) sinr, and
the bound state radial wavefunctions are given by p(r) = /(W,�,r) cosr -
g(W,�,r) sin r. (2.19) The hydrogenic requirement that the wavefunction be
finite atthe origin has been replaced by the requirement that at r > r0the
wavefunction be shifted in phase from the hydrogenic wavefunction by r.
This new boundary condition necessitates gfunctions as well as the regular /
functions inthe coulomb r>r0

Rydbergatom wavefunctions 19 region. We note that the wavefunction that we

generate is only good for r> r0. We have no knowledge about the Na
potential for r<r0 save the fact that it produces at r = r0 a wave phase shifted
from the analogous H solution by r, asis shown inFig. 2.6. The requirement
that ip �> 0 asr�> oo is equivalent to requiring that the coefficient of
u(v,�,r) be zero. This requirement leads to cos z sin Jtv + sin r cosnv = 0,
(2.20) or sin (nv + r) = 0.ThusJIV + r mustbe anintegral multiple of7t, and v
= n � tin where n is an integer. The effective quantum number v of a Na state
of angular momentum �differs from an integer by an amountT/JZ which we
recognize as the quantum defect d�. Since r isindependent of W, the quantum
defect (5�of aseries of �states isa constant. Due to the �(� + lyir1
centrifugal potential, inthe higher (, statesthe electron does not sample the
small rpart of thepotential as much, and the quantum defects are smaller. As
the radial wavefunctions for nonhydrogenic atoms obey the same r^> <*>
boundary condition they are similar to hydrogenic wavefunctions at large r,
but they do have different energies. Replacing the/radial function of Eq.
(2.15) bythe radial function of Eq. (2.19) leads to a wavefunction for a
nonhydrogenic atom valid for r > r0. Explicitly, ,r) cosjtd� - g(W,l,r) sin
jcde] ^ r which has the allowed energies where n is an integer. A point which
isworth keeping in mind isthat we have not considered core polarization
inobtaining thewavefunction of Eq. (2.21a). Due to the long range of a
polarization potential, it is not possible to define avalue of r0 such that for r
> r0 the potential is a coulumb potential. An important property of the
wavefunction is its normalization, and we have yet to normalize the radial
coulomb radial functions. Following the approach of Merzbacher, we can
find an approximate WKB radial wavefunction, good in the classically
allowed region, given by6 Nsin I' kdr' P(r) = 77= > (2-22) where N is a
normalization constant, r{ is the inner classical turning point and k = V2T, the
square root of twice the kinetic energy. The factor of Vifc in the denominator
reflects the fact that the amplitude of the wavefunction is larger where the
electron is moving more slowly. In the classically allowed region between
the inner and outer turning points, r{ and ro, ?_�C^l). (2.23)

20 Rydberg atoms We can use the wavefunction of Eq. (2.22) to obtain an

approximate normalization for the bound Rydberg states. Specifically we
require that C2JZ [IT Cr �V ' V^dr sin OdOdcj) = 1, (2.24) JO JO Jn where
rUo = n2�n Vn2 -�(� + l), (2.25) the plus and minus signs corresponding
to ro and rh respectively. Since the angular part of the wavefunction is
normalized, Eq. (2.24) reduces to ' p\r)dr=l. (2.26) Explicitly, Eq. (2.26) can
be written as sin2 f kdr'dr K2 r� , : ... �,, = !� (2.27) r2 Approximating
sin2 Jrr. kdr' by its average value 1/2 reduces the integral over r to one
which is easily done, with the result (2.28a) Tin or - < 2 2 S b ) The
normalization constant decreases as l/n3/2. At small r, the wavefunction's
only dependence on energy is through the normalization constant. For W> 0,
the Rydberg electron is no longer in a bound state but in the continuum. In
developing a wavefunction for the continuum we realize that the r-^> oo
boundary condition has been removed; only the r�> 0boundary condition
remains. For H, the r=0 boundary condition excludes the g function so
p(W,e9r)=f(WAr). (2.29) For Na or any other atom, the r �> 0 boundary
condition of H is replaced by the requirement that at r > r0 the wavefunction
is a wave phase shifted from the coulomb wave by r; p(W,�,r) = f(W,e,r)
cos r - g(W,(,r) sin r. (2.30) As r �� oo the wavefunctions are oscillatory
sine and cosine functions, as shown by Eqs. (2.14). For small r the
wavefunctions of the continuum are functionally identical to the bound
wavefunctions, differing only in their normalization. Since continuum waves
extend to r = oo they cannot be normalized in the same way asa bound state
wavefunction. We shall normalize the continuum waves per unit

Rydberg atom wavefunctions 21 energy. This is one of several options, and

we shall return to the reasons for this choice. If we again take advantage of
the fact that the angular part of the wavefunction is normalized, the continuum
normalization requirement is2 Too CW+AW pW)dr\ Pw,(r)dW' = l. (2.31)
JO JW-AW It is easier to work out the normalization integral for
normalization per unit wavenumber2 Ck+Ak p%{r)dr\ pk,(r)dk' = l. (2.32) 0
J k-Ak The normalization integral is dominated by the contribution from r^>
o�where k is a constant. To determine the normalization we use the WKB
form of a continuum wave, also given by Eq. (2.21). If wetake advantage of
the fact that the dominant contribution to the integral comes from large r,
where k is independent of r, we can express Eq. (2.32) as rk-Ak where we
have written the wavefunction normalized per unit wavenumber as (2.34) Nk
being the normalization constant to be determined. The equivalent
wavefunction normalized per unit energy is ^ (2.35) The integral of Eq.
(2.33) is easily reduced to Nl~=l- (2.36) The energy and wavenumber
normalization constants are related by the derivative dW/dk. Explicitly,2
N2k = N2w^=N2wk. (2.37) ak Using Eqs. (2.36) and (2.37) we calculate
Nw, yielding the wavefunction normalized per unit energy. Explicitly, Pw(r)=
p-sin \rkdr'. (2.38) ynk jri We note that this wavefunction differs from the
bound state wavefunctions by a factor of n~3/2. The factor of n~3/2 is, in
essence, the factor obtained by converting

22 Rydberg atoms from normalization per unit energy to normalization per

unit state. The conversion of the normalization integral requires the factor
dW/dn = d(-l/2n2)/dn = n~3. Explicitly, (^ N2wn~\ (2.39) Many of the
properties of Rydberg atoms depend on the wavefunction in the r > r0 region
where the potential is simply a coulomb potential. In this region we can very
easily calculate the wavefunction numerically using the Numerov method,
which can be applied to an equation of the form7 ^ = g(x)Y(x). (2.40) If x
increases in steps of size h, then the basic equation of the Numerov method
is7 [1 - T(x + h)]Y(x + h) + [1-T(x - h)]Y(x -h) = [2 + 10r(*)]Y(x) + O(h6),
(2.41) where T(x) = h2g(x)/12. If we ignore the terms of order h6 and if we
know Y(x) and Y(x � h) we can calculate Y(x + h), since everything else is
known. Zimmerman et al.8 calculated bound state radial functions and matrix
elements starting from Eq. (2.9). To remove thefirstderivative they made the
substitutions x = �n (r) (2.42a) and Y(x) = RVr, (2.42b) yielding a radial
equation in the form of Eq. (2.40) with g(x) = 2c2* (� - w) + ( � + | )2.
(2.43) In addition to transforming the radial equation to the Numerov form,
the transformation also changes the scale to one more nearly matching the
frequency of the oscillations of the radial function. For bound functions, the
wavelength of the wavefunction increases with r, and the substitution x =
�n(r) keeps the number of points per lobe closer to being constant. We start
in the classically forbidden region, r > 2n2, and decrease r to r ~ 0. We
follow this procedure since we know that as r�> �� the physically
interesting solution to the coulomb equation is a decaying exponential.
Successive values ofr are given by r}= rse~jh, where rs is the starting radius
(typically rs = 2n(n + 15)) and h isthe logarithmic step size, which isusually
chosen to be 0.01. As the index; is increased r decreases. As can be seen
from Eq. (2.41), we need to specify the values of Y at two adjacent points,
corresponding to / = 0 and 1. At r = rs the wavefunction is a factor of 1010
smaller than at the classical turning point, so a good estimate for Y(0),
corresponding to / = 0, is 10~10. The value of Y(h),
Rydberg atom wavefunctions 23 corresponding toy = 1, must only be larger
than the value of Y(0) to approximate the decaying exponential. The
orthogonal, nonphysical solution to the coulomb equation diverges as r�>
oo. Thus as we decrease r, toward r = 0, the wrong solution is exponentially
damped. The wavefunction can be calculated to the inner turning point or to
the radius of the ion corewhichever is encountered sooner. The calculation
must be stopped at this point, since for all nonzero quantum defects the g part
of the wave function will begin to exhibit its small r, r~~�(�+1) behavior
and diverge. The above procedure generates a wavefunction which is not
normalized. Its normalization integral is given by8 (2.44) With this
normalization integral it is possible to generate either a normalized
wavefunction, by converting Y(x) back to R(r), or matrix elements between
two bound wavefunctions. The Numerov technique described above assumes
that the outermost lobe is positive, which gives wavefunctions for n � I even
with the signs reversed from the usual convention that the functional form at
small r is independent ofn. Matrix elements of ra between two
wavefunctions, i.e. (llr0^), may be calculated by carrying out the Numerov
procedure for tworadial wavefunctions at the same time. The procedure must
be started at the appropriate starting point for the higher energy of the
twowavefunctions. For concreteness, let us say 11) has the higher energy.
The normalization integral N\ = XY\\<r\ is accumulated until the appropriate
starting point for the second wavefunction is reached. At this point we begin
to accumulate the second normalization integral N\ = 2Y2krk a s w e ^ a s the
matrix element R^2 � ^Y^^^C0 - The matrix element is thus given by8 12 An
alternative, physically appealing way of calculating the bound wavefunctions
wasused by Bhatti etal.9 In the classically allowed region, the kinetic energy
of the electron is given approximately by 1/r, thus the wavelength of the
wavefunction varies asVr. If wemake the substitution x = Vr and increment x
by a uniform step size /z, we should always have nearly the same number of
steps in each lobe of the wavefunction. Making the substitutions x = Vr
(2.46a) and Y(x) = r3/4R(r) (2.46b) leads to the following form of the radial
equation d2 dx x Y(x) = SY(x). (2.47)

24 Rydberg atoms In this case g(x) is given by g(x) =- 8 - Wx2 + 4J , (2.48)

and the solution has nearly the same number of points in each lobe of the
radial wavefunction. Using either of the above approaches it is
straightforward to calculate bound wavefunctions and bound-bound matrix
elements. Occasionally, though, we need continuum functions, which we
cannot generate in the same way. As r�� o� the physically significant
linear combination of/and g waves, i.e. /cos x � g sin r is not a decaying
exponential, but a sinusoidally varying wave phase shifted from the
hydrogenic / wave by r. Therefore the easiest way to calculate these
wavefunctions is to begin at the origin with an approximation to the regular /
function, which is proportional to r�+1, and calculate the wavefunction out
to a point where the coulomb potential is negligible compared to the kinetic
energy. At this point the wavefunction is simply a sine wave of constant
amplitude. Introducing a phase shift of r towards the origin produces the
desired coulomb wave at large r which may then be numerically propagated
to small r as r is decreased to zero.10 The wavefunction is normalized by
requiring that the amplitude of the sine wave be \Znl2k. Explicitly, we start
with the radial equation for p, Eq. (2.10), which is already in the Numerov
form. The Numerov substitutions for Eq. (2.40) which give the same number
of points per lobe as r �� o�are x = r (2.49a) and Y(x) = p, (2.49b)
yielding | ? j (2.50) Using the procedures outlined above we may calculate
bound and continuum wavefunctions as well as matrix elements of rCT, for
a^O, These wavefunctions are often called coulomb wavefunctions, and
properties calculated using them are said to be obtained in the coulomb
approximation. In addition, we can calculate matrix elements of inverse
powers of r for H. We cannot calculate with confidence matrix elements of
inverse powers of r for anything but H since the inverse r matrix elements
weight r ~ 0 heavily and the results can be highly dependent on the radius at
which we truncate the sums of Eq. (2.45). From their wavefunctions, we can
infer some of the properties of Rydberg atoms. We begin by estimating the
expectation values of r� where a isapositive or negative integer. The
expectation values of r�, a > 0, are determined mostly by the location of the
outer classical turning point, r = 2n2. Since the electron spends most of its
time there, a good estimate for the expectation values of r�, a > 0, is

Rydberg atom wavefunctions 25 Table 2.3. Expectation valuesofxafor H." =

�![5n2 = l/n2 1 ni(( + ni{� + -�(�� + 1 1/2) 1 ! ) ( � � + D] � 3
�(� H -h l/2)� H)] ;�+ i) 2n\t + 3/2)(� + 1)(� + l/2)�(� - 1/2) n 6, =
35^4 - 5n2[6�(�+ 1) - 5] + 3(1 + 2)(� + !)�(� - 1) V 7 8�v(�+ 5/2)
(� + 2)(� + 3/2)(� + 1)(� + l/2)�(� - l/2)(� - 1)(� - 3/2) a (fromrefs.
2, 11) Table 2.4. Properties of Rydberg atoms.0 Property n dependence
Na(lOd) Binding energy n~2 0.14eV Energy between adjacent n states n~3
0.023 eV Orbital radius n2 147 a0 Geometric cross section nA 68000 a02
Dipole moment (nd\er\nf) n2 143 ea0 Polarizability n1 0.21MHz cm2/V2
Radiative lifetime n3 1.0 jus Fine-structure interval n~3 �92 MHz "
(fromref. 1) {r�) ~ n2�. (2.51) If we are interested in inverse powers of r,
o < �1, it is the behavior of the small r part of the wavefunction which is
important, and therefore the precise value of the quantum defect is critical.
However, at small r, for n � �, the only energy, or v, dependence is on the
normalization, and thus (r-�) ocn~3 (2.52) While the n~3 scaling of the
expectation values of inverse powers of ris evident, the precise values are
not. For future use we give in Table 2.3 expectation values for H of (r�)
which have been determined analytically.211

26 Rydberg atoms Using combinations of radial matrix elements and energy

separations it is possible to deduce the n scaling of many properties of
Rydberg atoms. For example, the polarizability is proportional to the sum of
squares of electric dipole matrix elements divided by energy denominators,
and it is dominated by the contributions from a few nearby levels. The dipole
matrix elements between neighboring levels scale as the orbital radius, as n2,
while the energy differences scale asn~3, yielding polarizabilities that scale
asn7. Analogous reasoning allows usto develop scaling lawsfor many of the
properties of Rydberg atoms. Table 2.4 is a short list of representative
properties which shows several ways in which Rydberg atoms differ
substantially from ground state atoms. References 1. T. F. Gallagher, Rep.
Prog. Phys.51, 143 (1988). 2. H. A. Bethe and E. A. Salpeter, Quantum
Mechanics of One and Two Electron Atoms (Academic Press, New York,
1957). 3. M. Abromowitz and LA. Stegun, Handbookof Mathematical
Functions (Dover, NewYork 1972). 4. U. Fano Phys.Rev. A 2, 353 (1970). 5.
M. J. Seaton, Rep. Prog. Phys.46,167 (1983). 6. E. Merzbacher, Quantum
Mechanics (Wiley, New York, 1961). 7. J. M. Blatt, /. Comput. Phys. 1, 382
(1967). 8. M. L. Zimmerman, M. G. Littman, M. M. Kash, and D. Kleppner,
Phys.Rev.A 20, 2251 (1979). 9. S. A. Bhatti, C. L. Cromer, and W. E. Cooke,
Phys.Rev. A 24,161(1981). 10. W. P. Spencer, A. G. Vaidyanathan, D.
Kleppner, and T. W. Ducas, Phys.Rev.A 26, 1490 (1982). 11. B. Edlen, in
Handbuch der Physik (Springer-Verlag, Berlin, 1964).
3 Production of Rydberg atoms The original method of studying Rydberg
atoms was by absorption spectroscopy, and it remains a generally useful
technique. For example, the quasi-Landau resonances were first observed by
Garton and Tomkins using absorption spectroscopy.1 Many variants have
been developed in which a dye laser isused to populate an excited state
preferentially, allowing the absorption spectrum of the atoms in the excited
state to be recorded.2"4 However, most of the modern work on Rydberg
atoms has been carried out using methods in which the Rydberg atoms
themselves are detected. Rydberg atoms may be detected in two general
ways. First, since they are optically excited, they decay radiatively to low
lying states, emitting visible fluorescence, which is conveniently detected
using photomultiplier tubes. Second, in a Rydberg atom the valence electron
is in a large, weakly bound orbit, and as a result it is very easy to ionize the
atom by either collisional ionization or field ionization, after which either the
ion or the electron produced is easily detected. Excitation totheRydberg
states Rydberg atoms have been produced by charge exchange, electron
impact, and photoexcitation, the processes A+ + B ^ A n� + B+ (3.1a) e" +
A ^ A n� + e" (3.1b) hv + A->A nt, (3.1c) as well as by techniques in which
both collisional and optical excitation are used. Although combining
collisional and optical excitation has proven to be very effective, we shall
for the moment consider the three processes of Eqs. (3.1) separately. Charge
exchange has been used to convert ion beams to fast beams of Rydberg
atoms, and many of the techniques are described by Koch.5 Thermal beams
of Rydberg atoms have been made by electron impact or optical excitation of
beams of ground state atoms, and many of the relevant considerations are
described by Ramsey.6 It is also possible to produce thermal Rydberg atoms
in cells. Cells of glass or quartz are useful for containing permanent gases or
low densities (~10"5 Torr) of alkali atoms, and cells similar to optical
pumping cells 27

28 Rydberg atoms described by Happer7 have often been used in

fluorescence experiments. To contain higher densities (~ 1 Torr) of alkali or
alkaline earth atoms heat pipe ovens, as described by Vidal and Cooper,8 are
often used. The cross sections for all three of the processes of Eqs. (3.1)
scale asn~3, which may be easily understood. Consider, for example,
electron impact excitation of He. If an electron of energy Wo impingesupon a
ground state He atom it has atotal cross section, aI(W0), for exciting the He
atom from the ground state. Subsequent to excitation the final state of the He
atom is an excited He atom or a He+ ion anda free electron. The probability
of these outcomes is represented by daI(W0)/dW, where W is the energy of
the electron ejected from the He atom. W < 0 corresponds to the "ejected"
electron remaining in a bound state of He, and W > 0 corresponds to
ionization of the He. In general, do^/dWis asmooth function of W. More
important, there is no fundamental difference between W > 0 and W < 0. The
collision process occurswhen the electrons are within ashort range, <10A, of
the He+ core, and at these distances there is almost no difference between the
kinetic energies of a Rydberg electron and a free electron. As a result the
differential cross section doj/dW passes smoothly from W > 0 to W < 0. In
general oY depends on the energy, Wo, of the incident electron. However,
once Wo isfixedit is a good approximation to assume that daI(W0)/dWis
constant for W ~ 0 and given by (3-2) Below the limit, W < 0, the energy is
no longer a continuous variable, and it is more useful to write the cross
section per principal quantum number a(�, Wo) = � = � � = �> (3.3) dn
dW dn n showing the n 3 dependence of the cross section for the production
of Rydberg states. Another way of seeing the nscalingis the following. Prior
to the collision the He atom is compact, having a radius of approximately 1
A. Immediately after the collision the Rydberg electron isin its large orbit,
which has adensity at the origin scaling asri~3.Thus when the initial state is
projected onto the final state in the excitation process, it is hardly surprising
that the cross section has an n~3 dependence. Similar arguments apply to
charge exchange and photoexcitation, and the basic result is the same; the
cross section for the production of Rydberg atoms is the continuation below
the limit of the ionization cross section, leading to an n~3 dependence of the
excitation cross section.

Electron impact 29 ^DETECTION f GRIDS kV/cm ELECTROSTATIC

HIGH RYDBERGS ANALYZER PLATES 0-l5kV/cm 0 2 4cmSHIELD.
MAGNETIC EXTRACTOR l5V/cm 8 SHIELD Fig. 3.1 Apparatus for
electron beam excitation of rare gas atoms (from ref. 9). Electron impact
Schiavone et al. have measured the cross sections for the production of
Rydberg states of rare gases using the apparatus shown in Fig. 3.1.910
Rydberg states of rare gas atoms are formed by electron impact by a pulsed
electron beam in the sourcefilledwith 10~5Torr ofrare gas. Some ofthe
Rydberg atoms formed travel towards thedetector, which ionizes allatoms
which can beionized by a 12 kV/cm field and detects the resulting ions. The
distribution of Rydberg states reaching the detector is affected by several
processes, theexcitation from the ground state, � changing collisions
producedby the electron beam, and radiative decay ofthe Rydberg atoms
during theflightfrom the exciting electron beam to the detector. To extract
theexcitation cross section they first measured the number of Rydberg atoms
reaching the detector as a function of ionizing field byvarying the voltage on
the analyzer plates shown inFig. 3.1.The observed Rydberg atom signals as a
function offieldwere converted to ndistributions by making several
assumptions. First they assumed that the excitation cross section scales as
n~3. Second, they assumed the radiative lifetime to be given by r^3, r1 being
a constant. Finally, ionization was assumed to occur at the classical field,
E=l/16n\ (3.4) Using these assumptions they could then fit their data to
determine excitation cross sections, a(n, Wo), forthe excitation ofaRydberg
state of principal quantum number n by an electron of energy Wo. They found
the values of o[(W0) given in

30 Rydberg atoms Table 3.1. Electron impact cross sections for the
excitation of rare gas Rydberg statesfor the electron energy atthepeak of the
cross section and 100 eV.a Rare gas He He Ne Ne Ar Ar Kr Kr Xe Xe fl
(fromref. Electron energy Wo (eV) 70 100 60 100 28 100 20 100 20 100 10).
(A2) 0.77 0.67 0.63 0.61 6.5 1.5 4.0 2.0 10.0 4.6 Table 3.1 for the rare
gasatoms.10Our o[(W0) corresponds to oex(n = 1, E) in the notation of
Schiavone etal.10. For atypicalvalueof o'(Wo) � 1 A, thecross section a(20,
Wo) for exciting an n = 20 state is 1.25 x 10~4A2. An interesting aspect of
electron impact excitation noted bySchiavone etal.10is that the electrons play
two roles. First, they excite the ground state atoms to Rydberg states. Second,
they collisionally redistribute the atoms initially excited to low � states
over all � states, including the long lived high � states. The production of
high �states allows the detection of n states which would normally not live
long enough to reach the detector. The cross section for � changing
collisions by electrons is so large, ~106 A2 , that only at low currents does
the observed signal depend on the electron current in an obviously nonlinear
way. At higher electron currents the �mixing isso saturated that the signal
islinear in the electron current. Electron impact excitation has alsobeen used
in along series of experiments by Wingand MacAdam11to measure the
A�intervals of He Rydberg states by radio frequency spectroscopy. The fact
that there are observable population differences makes it clear that electron
collisions do not produce a thermal population distribution on the time scale
of the radiative decay of the Rydberg states in these experiments. In sum,
electron impact excitation has the advantages of relative simplicity and
generality, and it hasthe disadvantages of being inefficient and
nonselective,with nearly all energetically possible states being produced.

Charge exchange 31 ion ^ source r H+ charge exchange ion deflector Hni H+

. labelling ��. field Fig. 3.2 The fast beam approach of Bayfield and Koch
(ref. 13).H+ ionsof roughly 10 keV energy pass through a charge exchange
cell forming a fast beam of H Rydberg atoms. Down-stream from the charge
exchange cell the ions are deflected from the beam and a band of n states
isselected by a square wave modulated ionization field. Charge exchange
Charge exchange of H+ and He+ with various target gases has been used to
produce fast beams of Rydberg atoms since the first experiments of Riviere
and Sweetman.12 The method of producing fast Rydberg H atoms used by
Bayfield and Koch13 in their initial experiment to study the microwave
ionization of Rydberg atomsis shown schematically inFig. 3.2. The 10
keVprotons from an ion source pass through a charge exchange cell filled
with Xe, where some of the protons are converted to fast Rydberg atoms. The
distribution of product states of charge exchange of ions with incident energy
Wo is described by da^(W0)/dW, where W is the energy of the Rydberg
electron in the neutral product atom and crL(Wo) is the total electron loss
cross section. As in Eq. (3.3), we can write the charge exchange cross
section for the population of a specific n state as (3.5) n where oi(W0) =
daL(W0)/dW|w=0. In Table 3.2 we give representative values of
a^(Wo).14"16 Only asmall fraction of the incident ions which undergo
collisions with the gas in the charge exchange cell will be left in any
particular Rydberg state. For n = 10 the fraction is <10~3, and for n = 30it is
<10~4. If the number density of the charge exchange gas is N and the length
of the cell is L, in principle the charge exchange cell can be operated with
NLo^ ~ 1, so that nearly all the ions in the beam undergo charge exchange. In
fact it is usually not possible to use such a thick target since it will scatter the
fast neutral atoms into large enough angles to remove them from the beam.
Since the cross section for scattering atoms out of the beam is ~ 1 A2, the
cell must be operated at a density length product low enough that the Rydberg
atoms produced are not scattered out of the beam. The Rydberg
statesproduced inthe cellbycharge exchange can be collisionally depopulated
by subsequent collisions with the target gas. In the charge exchange collision,
it is presumably low � states which are populated, by virtue of the overlap
oftheir wavefunctions withtheground statewavefunction attheorigin.In

32 Rydberg atoms Table 3.2. Values of the charge exchange crosssectionfor

protons on severalvapor targets. Target Xefl Kr" Ar" Ar* He* Hec H / N2C
CO2C Incident energy (keV) 20 20 20 30 20 60 60 60 60 (A2) 10 9 4 3.3 0.4
0.6 1.1 2.5 3.8 "(seeref. 14) ^(seeref. 15) c (see ref. 16) thermal collisions it
has been observed that the cross sections for �and m changing collisions are
large,1718 with values for Xe reaching 105 A2 at n = 20, and minimum
values >20 A2. These cross sections are far larger than crL, so it is likely
that efficient charge exchange is accompanied by collisions which populate
high � and m states. Thermal energy cross sections for changing n are
smaller than the � changing cross sections, with typical values of �10 A2.
Since virtually all n states are populated, some further redistribution over n
has no significant effect. In sum it is reasonable to expect a distribution of
n�m states from charge exchange, with n~3 scaling of the population in each
n. Such adistribution of states must be filtered in some way before it is useful
in an experiment. A good way is the method used by Bayfield and Koch and
shown in Fig. 3.2.13 The Rydberg atoms pass between twofieldplates, which
produce a modulated field to select a band of n states by field ionization. The
field required to ionize a state of principal quantum number n is given by an
expression similar to Eq. (3.4). If the field is switched between Ex and E2
with E2< Ex, atoms in states of n > nt or n2 are ionized by the field. Thus the
difference in signal obtained with E = Ex and E = E2 must be due to atoms
with nx<n<n2. Using a field switched between 28.5 and 41.0 V/cm, Bayfield
and Koch selected the band 63 < n < 69 for the first experiments with
microwave ionization.13 Optical excitation Optical excitation differs from
collisional excitation in a fundamental way; the exciting photon is absorbed
by the target atom. As a result, specifying the energy

Optical excitation 33 -300 -200 -100 W (cm"1) Fig. 3.3 Excitation spectrum
of the Na ns and nd states from the Na 3p3/2 state (from ref. 19). of the
absorbed photon specifies the Rydberg state produced. In contrast, specifying
the energy of an incoming electron does not specify the energy of the Rydberg
state produced because there is no way to control how the energy is shared
between the incident electron and the electron which is excited to the
Rydberg state. In spite of the above difference from collisional excitation,
photoexcitation has the similarity of having across section with an n~3
dependence. Thiscross section is simply the continuation of the
photoionization cross section from above the limit, W >0,to belowthe limit,
W < 0.Thispoint is demonstrated experimentally by Fig. 3.3, the excitation
spectrum of the Na ns and nd states from the 3p3/2 state.19 The signal, or the
cross section, is apparently the same just above and just below the limit, in
spite of the fact that the electron isphotoionized in the former case and
excited to a bound Rydberg state in the latter. Only when n < 40is the
instrumental resolution adequate for resolved Rydberg series to appear in the
spectrum, and for n < 30 the ns and nd Rydberg states are clearly resolved
and increase inintensity with decreasing n. A useful wayto
describephotoexcitation is to specify oPi, the photoexcitation cross section in
the vicinity of the ionization limit, W = 0. Above the limit the photoionization
cross section is given by crPI. Below the limit, the cross section averaged
over an integral number of n states equals crPI. The crosssection a(n)for
exciting a resolved Rydberg stateis given by

34 Rydberg atoms Table 3.3. Photoionization cross sections at threshold (W

= 0).a Wavelength aPI Atom (A) (10"18 cm2) H Li Na K Rb Cs Mg Ca Sr fl
(fromref. 912 2299 2412 2856 2968 3184 1621 2028 2177 20) a(n) = 6.3
1.8 0.125 0.007 0.10 0.20 1.18 0.45 3.6 tfpi AWnr (3.6) where AWis the
energy resolution of the excitation. The cross section for exciting the Rydberg
state is the photoionization cross section times the ratio of the An energy
interval to the experimental resolution. In most optical excitations the
resolution is determined by the Doppler effect or the finite linewidth of the
light source. The Doppler effect gives a typical frequency width of 1GHz,
and the width of the light source can be anywhere from 1 kHz to 30 GHz. We
assume that these widths are larger than the radiative width. The
photoionization cross sections from the ground states of H, alkali, and the
alkaline earth atoms are given in Table 3.3. 20 The most widely used method
of optically exciting Rydberg atoms is to use a pulsed dye laser pumped by a
N2 or NdrYAG laser. It is straightforward to generate visible laser pulses
5ns long with 100juJ energies and 1 cm"1 bandwidths. Such a pulse contains
~3 x 1014 photons, and, if collimated into a beam of cross sectional area
10~2 cm2, it has an integrated photon flux of 1016 cm"2. At n = 20 the An
spacing is �28 cm"1, and with aresolution of 1 cm"1, the cross section of
Eq. (3.6) is a factor of 28 larger than the photoionization cross section. Using
10~18 cm2 as the photoionization cross section we find a probability of 10%
for exciting atoms exposed to a photon flux of 1016cm"2 to the n = 20 state.
A typical example of the pulsed dye laser excitation is the beam experiment
shown in Fig. 3.4 in which Na atoms in a thermal beam are excited in two
steps from the ground state 3s to the 3p state with a yellow dye laser photon
and from the 3p state to a high lying ns or nd state with a second, blue
photon.21 Using continuous wave (cw) laser excitation it is possible to
excite atoms with substantially higher efficiency than using pulsed lasers. For
example a single mode laser of 1 MHz linewidth has a resolution 3 x 104
better than the pulsed laser

Collisional-optical excitation 35 Fig.3.4Apparatusfor thestudyof

Rydbergstatesof alkalimetal atoms:a,theatomicbeam source; b, the
electricfieldplates;c,the pulsed laser beams; and d, the electron multiplier
(fromref. 21). described above. Accordingly, for single photon excitation the
resolution limited cross section can be 3xlO4 times larger if the excitation is
done so as to avoid Doppler broadening. Examples of such an excitation are
the experiments of Zollars et al.22 using a frequency doubled cw dye laser to
excite a beam of Rb ground state atoms to the np states and the experiments
of Fabre etal23 in which they used a cw dye laser and diode lasers to excite
Na nf states. Collisional-optical excitation Purely optical excitation is
possible for alkali and alkaline earth atoms. For most other atoms the
transition from the ground state to any other level is at too short a wavelength
to be useful. To produce Rydberg states of such atoms a combination of
collisional and optical excitation isquite effective. A good example isthe
study of the Rydberg states of Xe by Stebbings etal24 As shown in Fig. 3.5, a
thermal beam of Xe atoms is excited by electron impact, and a reasonable
fraction of the excited atoms is left in the metastable state. Downstream from
the electron excitation the atoms in the metastable state are excited to a
Rydberg state by pulsed dye laser excitation. Metastable atoms have
alsobeen used as the startingpoint for laser excitation in cell experiments.
Devos et al.25 used metastable He atoms in the stationary
36 Rydberg atoms Laser beam | in collector \ Fused quartz window^ ^ v ^ ^ z
~ S u r f a c e detector Mirror -Charged particle j ^ deflector Vacuum wall
Neutral beam source Fig. 3.5 Schematic diagram of the apparatus using
electron impact excitation to theXe metastable state followed bylaser
excitation to Rydberg states (from ref. 24). afterglow of a pulsed discharge
as the starting point for pulsed laser excitation to high lying np states. They
observed the time and wavelength resolved fluorescence to determine the
collision rate constants of the np states. An example in which selective
detection of the Rydberg atoms isnot required isthe optogalvanic
spectroscopy of Ba Rydberg states of Camus et al.26 They optically excited
the atoms from metastable levels produced in a discharge to high lying
Rydberg states. The Rydberg atoms are collisionally ionized and the resulting
ions and electrons alter the discharge current, a readily detectable signature
of the excitation of a Rydberg state. In fast beams optical excitation has
proven to be most useful. Since the fast beams are low in intensity, but
continuous, cw lasers have been used. Usually, fixed frequency lasers have
been used sincefinetuning can be done using the Stark shift or the Doppler
shift of the fast beam. The Doppler shift can be used either by changing the
angle at which the laser beam and fast beam cross, or by altering the velocity
of the fast beam. An early example wasthe use of the uvline of anAr laser to
drive transitions from the metastable H 2s state to the 40 < n < 55np states.27
In thisparticular case the velocity of the beam waschanged totune different np
states into resonance. The most commonly used technique has been to use a
CO2 laser which has about 20 lines at intervals of 1.3 cm"1in each of the 9.6
and 10.6jumbands. These lines match the transition frequencies of n = 10to n
~ 30transitions,28 and allow efficient selective population of single states of
n ~ 30. The essential notion is to

References 37 populate the Rydberg levels of the fast beam by charge

exchange, remove all atomsinn > 10 statesbyfieldionization, and then
repopulate the selected level of n ~ 30 usingthe CO2laser. Thisapproach, or
avariant usinga second CO2 laserto drive n =7to n = 10 transitions, has
become a standard approach in the studyof Rydberg atoms using fast
beams.28 References 1. W. R. S. Garton and F. S. Tomkins,Astrophys. J. 158,
839 (1969). 2. D. J. Bradley, P. Ewart, J. V. Nicholas, and J. R. D. Shaw,/.
Phys.B 6, 1594,(1973). 3. J. R. Rubbmark, S. A. Borgstrom, and K.
Bockasten, / Phys.B 10,421 (1977). 4. E. Amaldi and E. Segre, Nuovo
Cimento 11,145 (1943). 5. P. M. Koch, in Rydberg States of Atoms and
Molecules, eds R.F. Stebbings and F.B. Dunning (Cambridge University.
Press, Cambridge, 1983). 6. N. F. Ramsey, Molecular Beams (Oxford
University Press, London, 1956). 7. W. Happer, Rev. Mod. Phys.44,169
(1972). 8. C. R. Vidal and J. Cooper, /. Appl. Phys.40, 3370 (1969). 9. J. A.
Schiavone, D. E. Donohue, D. R. Herrick, and R. S. Freund, Phys.Rev.A 16,
48 (1977). 10. J. A. Schiavone, S. M. Tarr, and R. S. Freund, Phys.Rev. A
20,71 (1979). 11. W. A. Wing and K. B. MacAdam, in Progress in Atomic
Spectroscopy A, eds W. Hanle and H. Kleinpoppen (Plenum, New York,
1978). 12. A. C. Riviere and D. S. Sweetman, in Atomic Collision
Processes, ed. M. R. C. McDowell (North Holland, Amsterdam, 1964). 13.
J. E. Bayfield and P. M. Koch, Phys.Rev. Lett.33, 258 (1974). 14. R. F. King
and C. J. Latimer, /. Phys B. 12, 1477 (1979). 15. J. E. Bayfield, G. A.
Khayrallah, and P. M. Koch, Phys.Rev.A 9, 209(1974). 16. R. N. W in, B.
Kikiani, V. A. Oparin, E. S. Solov'ev, and N. V. Fedorenko, Sov. Phys. JETP
20, 835(1965). [/. Exptl. Theor. Phys. USSR 47, 1235 (1964)]. 17. R.
Kachru, T. F. Gallagher, F. Gounand, K. A. Safinya, and W. Sandner,
Phys.Rev. A 27, 795(1983). 18. M. Hugon, F. Gounand, P. R. Fournier, and J.
Berlande, /. Phys.B 12,2707 (1979). 19. W. R. Anderson, Q. Sun, and M. J.
Renn private communication. 20. G. V. Marr, Photoionization Processes in
Gases (Academic Press, New York, 1967). 21. D. Kleppner, M. G. Littman,
and M. L. Zimmerman, in Rydberg States ofAtoms and Molecules, eds R. F.
Stebbings and F. B. Dunning (Cambridge University Press, Cambridge,
1983). 22. B. G. Zollars, C. Higgs, F. Lu, C. W. Walter, L. G. Gray, K. A.
Smith, F. B. Dunning, and R. F. Stebbings, Phys. Rev. A 32, 3330(1985). 23.
C. Fabre, Y. Kaluzny, R. Calabrese, L. Jun, P. Goy, and S. Haroche, /. Phys.
B 17, 3217 (1984). 24. R. F. Stebbings, C. J. Latimer, W. P. West, F. B.
Dunning, and T. B. Cook, Phys.Rev.A 12, 1453 (1975). 25. F. Devos, J.
Boulmer and J. F. Delpech, /. Phys. (Paris) 40, 215(1979). 26. P. Camus, M.
Dieulin, and C. Morillon, /. Phys. Lett. (Paris) 40, L513 (1979). 27. J. E.
Bayfield, L. D. Gardner, and P. M. Koch, Phys. Rev. Lett.39, 76 (1977). 28.
P. M. Koch and D. R. Mariani, /. Phys.B 13, L645 (1980).25

4 Oscillator strengths and lifetimes In Chapter 3we considered briefly the

photoexcitation of Rydberg atoms, paying particular attention to the continuity
of cross sections at the ionization limit. In this chapter we consider optical
excitation in more detail. While the general behavior is similar in H and the
alkali atoms, there are striking differences in the optical absorption cross
sections and in the radiative decay rates. These differences can be traced to
the variation in the radial matrix elements produced by nonzero quantum
defects. The radiative properties of H are well known, and the radiative
properties of alkali atoms can be calculated using quantum defect theory.
Oscillator strengths A convenient way of expressing the strengths of
transitions is to use the oscillator strength. The oscillator strength
fn<z>m>,nzm from level nim to level n'l'm' is defined by1 h'Vm'Mm =
2ja)n^9J (n't'm \x\Mm) |2, (4.1) where a)n^> n� = (Wn>�> � Wn�)/h. We
could equally well have used the matrix elements of y or zin Eq. (4.1), but in
any ca.se,fn>e>m>jW�mis m dependent. Since it is evident that the
radiative decay of an atom in free space cannot depend upon m, it is useful to
introduce the average oscillator strengthfn>^^ which does not depend upon
m. It may be written as1 where �max is the larger of � and V'. If wereverse
the roles of � and �', it is straightforward to show that 2�'+ l n�,n'�'-
(4.3) The usefulness ofthe oscillator strengths stems in part from the fact that
they satisfy several sum rules. The Thomas-Reiche-Kuhn sum rule is given
by1 fn'i'm'Mm = Z, (4.4) where Z isthe number of electrons in the atom and
the sum implicitly includes the accessible continua. When applied to Na, this
sum rule leads to ^fn'z'm',ntm = H> 38

Oscillator strengths and lifetimes 39 which tells us nothing about the

oscillator strength of transitions of the outer electron alone. More useful
sumrules,validfor oneelectron in a central potential, are1 and (4.6) from
which it is apparent that "7�-Orf = l, (4-7) n't' where �' = ��1. Eqs.
(4.5)-(4.7) apply to, for example, the transitions of the valence electron of
Na, in spectral regions where the transitions of the Na+ core are unimportant.
Since the lowest lying excited states of Na+ lie approximately 25eV above
the ground state of Na+, and 30 eV above the ground state of Na, Eqs. (4.5)-
(4.7) can be used to described the transitions of the valence electronof Na
due to photons of energy less than 10 eV. The attractions of the oscillator
strengths are that they are dimensionless and they sumto 1. Thusthe
determination of one oscillator strength is often enough to give a good
indication of the overall distribution of oscillator strengths. We shall see that
the effect of nonzero quantum defects is to redistribute the alkali oscillator
strengths relative to those of H. If we return to Eqs. (4.5) and (4.6)we can
see that the strongest ��� �-1 transitions are to lower energies and that
the strongest �-� �+1 transitions are to higher energies. Finally, we
introduce the Einstein A coefficient, An'Vni, which defines the spontaneous
decay rate of the n� state to the lower lying n't' state. Explicitly,1 An>i, ne =
- � " y ^ax \(n>t>\r\nt)\2. (4.8) n i M 3hc3 2M-11 In terms of the average
oscillator strength we can write the A coefficient as An'�',ne =
Theminussigncomesfrom thefact that/^�' n^> < 0.The radiative lifetime,
rn�, of the n( state is simply the inverse of the total radiative decay rate,
which is obtained by summing An>z>n� over all lower lying n'V states.
Explicitly (4.10) From Eq. (4.8) we see that the A values contain a factor of
&>3, which generally means that the transition with the highest frequency
contributes the most to the radiative decay rate and therefore dominates the
overall dependence on n.

40 Rydberg atoms For all Rydberg n� states savethe sstates the highest
frequency decay transition is to the lowest lying state of �- 1. For an sstate it
isthe transition to the lowest p state. In either case, for a high n� state, as
n�� o� the frequency of the highest frequency transition approaches a
constant. Thus, in the limit of high n, the A value depends only on the radial
matrix element between the Rydberg state and the low lying state. Only the
part ofthe Rydberg state wavefunction which spatially overlaps the
wavefunction of the low lying state contributes to the matrix element in Eq.
(4.8). As a result, the squared radial matrix element of Eq. (4.8) exhibits an
n~3 scaling due to the normalization of the Rydberg state wavefunction.
Correspondingly, rneocn\ (4.11) For �~n states the same reasoning does not
apply. For the � = n� 1 state the only allowed transition is to the n' = n �
1, (,' = n � 2 state. In this case the frequency of the transition is not constant
but is given by 1/n3. When cubed, this frequency contributes ann~9 scaling
tothe decay rate. Offsetting this scaling is the fact that the radial matrix
element represents the size of the atom in the n and n � 1 states. Since (n �
In � 2\r\n n � 1) ~rc2thelifetimes of the highest � states scale as n5, i.e.
rnt-n-i"*5' (4-12) The average lifetime for a statistical mixture of �rastates
of the same n is given by1'2 rnccn4-5. (4.13) Finally, itis useful to extend the
definition of oscillator strength above the ionization limit using = ? _ ^ L \
(�e> \r\n{)\2 (4.14) where the continuum wavefunction is normalized per
unit energy. The photoionization cross section of a statistical mixture of m
levels of the state to the e'�' continuum is then given by cdW (4.15) It is
interesting to examine the dependence on n and (of the oscillator strength and
lifetimes for the H and alkali atoms. First we show in Fig. 4.1 a plot of the
oscillator strengths of the H Is -�rip transitions using the format of Fano and
Cooper.3 Above the ionization limit df/dW is plotted directly. Below the
ionization limit the oscillator strength to the np level isplotted asarectangular
blockof area/np,lsfrom W= -l/2(n - l/2)2to W = -V2(n + 1/2)2,i.e. it is
approximately 1/n3 wide,/np lsn 3 high, and centered at energy W = - l/2n2.
Thus the area of each block corresponds to the oscillator strength for the
excitation of one np state. Plotting the oscillator strength in this fashion
brings out clearly the fact that

Oscillator strengths 41 -0.2 -0.1 0 0.1 W(a.u.) Fig. 4.1 Oscillator strength
distribution from the H Is state to the excited p states. Below the ionization
limit the area of each block corresponds to the average oscillator strength
from the Is state to an np state. oscillator strength per unit energy passes
smoothly from the bound to the continuous region of the spectrum. From Fig.
4.1 it is also apparent that at the limit the oscillator strength is slowly
varying, and thus fnp ls ^ n~3 as expected from the normalization of
thenpwavefunctions near theorigin.Fig.4.1 and direct calculation show that a
reasonable fraction, about one half, of the oscillator strength lies above the
ionization limit. In the excited states of H the lifetimes exhibit apredictable
behavior. For states of agiven nthe p states decay most rapidly, followed
bythe d,f,g,s,h . . .statesin that order.4 Putting aside the s states, the dominant
decay mechanism for each nl series is to the lowest n'� � 1 state. The
frequency dependence of this transition upon �leads to the observed
�ordering of the lifetimes. The ns states have long lifetimes due to the fact
that the radial matrix elements to the lower lying p states are small. For the
high ns states of H the first two lobes of the ns radial wavefunction overlap
the 2pwavefunction, andtheir contributions canceltosome extent. In contrast,
only one lobe, the innermost, of the nd radial wavefunction overlaps the 2p
radial wavefunction. The nd wavefunction also has a larger amplitude than
the ns wavefunction at the position of the 2p wavefunction because of the
centrifugal potential, which ensures that the nd atom has less kinetic energy
inradial motion. In general,matrix elementsinvolvingdecreases in n
accompanied by a decrease in � byone arelarger than matrix elements
involving

42 Rydberg atoms Fig. 4.2 (a) Radial matrix element of the K 4s->rc*p
transition with n* a continous variable. The quantum defect of the 4s state
is2.23. The quantum defects of the np states are 1.71, so the np states fall at
the locations shown by the arrows near where the matrix element crosses
zero, (b) H 2s-rc*pradial matrix element asafunction of n*. Note that the
maximum amplitudes of the matrix element occur at integer values of �*.
decreases in nwith an increase of �byone. With the exception of the ns
states, all states decay predominantly by A�= -1transitions.1'4 Now let us
consider alkali atoms. We begin by considering the K 4s-np excitations
analogous to the excitation from the ground state of H. In K the situation is in
fact quite different, as shown by Fig. 4.2(a), a plot of the radial

Oscillator strengths 43 � D 10"2 4p 5P 6p 7p W(a.u.) Fig. 4.3 Oscillator

strength distribution from the K 4s state to the np states and �p continuum.
Note that the vertical scale is logarathmic, not linear as in Fig. 4.1. The
oscillator strengths to the high np states are orders of magnitude smaller than
in H. matrix element (rc*p|r|4s) for K with n* a continuous variable. The K
4s state has a quantum defect of 2.23, and the high lying p states have
quantum defects of 1.71, sothat ?z* = 0.29(mod 1). In Fig. 4.2 we have
shown the locations of the npstates. It isquite apparent that the 4s �>
4ptransition has alarge matrix element, but the matrix elementstohigher n
statesfall near thezerocrossingsinthematrix element and are therefore very
small. The analogous plot of the matrix element between the H 2s and np
states is shown in Fig. 4.2(b). The H (2p|r|2s) matrix element is larger than
the K 4s^p matrix element due to the smaller binding energy of the H 2s state,
and all the matrix elements between the 2s and the higher np states are further
away from the zero crossings and hence larger than the K(np|r|4s) matrix
elements. If we convert the squares of the matrix elements shown in Fig. 4.2
to average oscillator strengths, by multiplying by 2a>/3, we find that in H the
2s ^2p transition has vanishing oscillator strength due to the 2s-2p
degeneracy. All the oscillator strength must come from the n > 3 2s-> np and
2s -> ep transitions. On the other hand, in K the 4s�> 4p transition has a
significant oscillator strength sincethe 4p state lies halfway between the 4s
state and theionization limit.In fact, it has an oscillator strength of very nearly
1, consuming nearly all of the available K4s-np oscillator strength. In Fig.
4.3 we show aplot of d//dWfor K. The values for the bound oscillator
strengths and for the continuum more than 0.03 au above the limit are taken
from Marr and Creek.5'6 The values of df/dW near the minimum just above
the limit are taken from the relative measurements of Sandner etal.7
normalized to the bound state oscillator strengths given by Marr

44 Rydberg atoms and Creek.5 In K the variation in d//dW is so great that

Fig. 4.3 is plotted on a logarithmic scale. As shown, df/dW declines
throughout the bound spectrum reaching a minimum just above the ionization
limit. This minimum, usually termed the Cooper minimum, is found in the
continua of all the alkalis save Li. The origin of the minimum ismost
easilyunderstood by considering aspinless K atom. In a spinless K atom the
4s-np radial matrix elements decrease with increasing energy, as implied by
Fig. 4.3.The decrease with increasing energyof the np state occurs because
the nodes of the bound np, or continuum �p, wavefunction become
increasingly close together with increasing energy. Stated another way,the 4s-
np or4s-�p radial matrix element is anintegral of the product of a fixed
function, r3R4s(r), and a function the wavelength of which decreases with
energy, Rnp(r) or R�p(r). In general the resulting integral oscillates about
zero as a function of energy. In K the first change in sign is just above the
ionization limit, leading to d//dW = 0 and a vanishing photoionization cross
section. Having neglected spin we predict a zero cross section instead of the
nonzero minimum of Fig. 4.3. The energy at which the minimum occurs is
determined by the radial phases of the ground state and ep wavefunctions. In
Na, K, Rb, and Cs these phases differ by ~yr/2,sincethe ground state and
npquantum defects differ by 1/2. For these atoms the Cooper minimum occurs
in the continuum not far above the ionization limit. In Li the quantum defects
of the ground state and the np states differ by 1/4, and d//dWhas itsminimum
below the ionization limit. As shown by Fig. 4.3, the cross section does not
actually vanish, but only decreases to anonzero minimum. As pointed out by
Seaton,8when the spin orbit interaction is taken into account there are two ep
continuum waves, ep1/2 and �p3/2, which have slightly different radial
phases corresponding to the slightly different quantum defects of the bound
np1/2 and np3/2 series. Asa result of having different phases the epV2 and
ep3/2 radial matrix elements cross zero at slightly different energies. Each of
these continuum channels has its own oscillator strength, and the total cross
section is never zero. While the location of the minimum isa function of the
difference inthe quantum defects of the ground state and the excited np states,
the minimum cross section and the width of the minimum are proportional to
the np fine structure splitting, or the difference in the np1/2 and np3/2
quantum defects. Consequently the minimum crosssection and the width of the
Cooper minimum increase from Na to Cs. Since the radial matrix elements
from the ground state to the npV2 and np3/2 states are not the same, due to the
differing radial phases, the oscillator strengths and intensity ratios of the
transitions to the np1/2 and np3l2 states cannot be in the 1:2 ratio predicted
by the angular factors.8 The largest deviations occur in Cs, for the Cs 6s1/2-
�pi/2 matrix element vanishes just above the ionization limit.9 As a result,
just below the ionization limit large intensity ratios have been
measured.10"13 For example, atft=30Raimond etal.measured a ratio of
1:1170.13 The measurements agree with calculations of Norcross14 which
predict that the

Radiative lifetimes 45 intensity ratio atn= <*> should be roughly

1:12,000.While the Cooper minimum in Kfalls much closer to the ionization
limit than it does in Cs, both the 4s-�p1/2and 4s-�p3/2 matrix elements
cross zero at nearly the same energy, andin the bound states themaximum
wpy2:np3/2 intensity ratio isonly 1:4.15 Radiative lifetimes Both the
absorption cross sections and the lifetimes oflow�states ofalkali atoms are
different from the hydrogenic values. The most glaring examples occur the
alkali np states. In H the np states have very short radiative lifetimes, but in
an alkali atom they have very long lifetimes, due to their small oscillator
strengthsto the ground state. For example, as shown in Figs. 4.1 and 4.3, the
oscillator strengths from the ground states of H and K to the high, n ~
10,npstates differ by a factor of 100. This factor, when multiplied by the ratio
of the squares of the frequencies, 16, leads to K np^> 4s Einstein A
coefficients roughly 1000 times smaller than their H np�� Iscounterparts.
The Knp lifetimes are nota thousand timeslonger than the H nplifetimes
because decay to states other than the ground state plays a significant role in
thetotal decay rate. The radiative lifetimes of many excited states of
Li,16Na,17"21 K,22 Rb,23"25 and Cs,26"28 have been measured using a
variety of techniques, the most common beingtime resolved laser
inducedfluorescence,whichis typically carried outusing a cell, asshown in
Fig. 4.4.In allcases, the observed lifetimes arein reasonable agreement
withvaluescalculated in the coulomb approximation, corrected for the
decrease dueto black body radiation.29 In thefollowing chapter we show
thatif the 0 K lifetime of a state is r that its lifetime at a finite temperature T
is given by24'30 i = ; + i- (4.16) where l/rbb is theblack body radiation
induced decay rate. Systematic lifetime measurements have been made for
Rb23"25 and Na.17"21 In both cases the lifetimes of the s, p, d, and f states
have been measured, and we discuss the Na lifetimes as arepresentative
example. The lifetimes of Nan < 15 s, p, d, and f states have been measured
using the time and wavelength resolved resolved laser
inducedfluorescencearrangement shown in Fig. 4.4. Na atoms at a pressure of
10~6 Torr in apyrex cell are excited from ground state tothe 3p state and then
tothe ns ornd states using two dye laser pulses of5 ns duration. Thens and nd
lifetimes are easily measured by observing the time resolved ns�> 3p or
nd^> 3pfluorescencewith aphotomultiplier1718. These fluorescent decays
have largebranching ratios and areeasilydetected inspiteofthefact that they
are atthe

46 Rydberg atoms laser beams heated oven Fig. 4.4 Schematic diagram of a
typical arrangment for laser induced fluorescence measurements of lifetimes.
A pulsed laser beam (or beams) passes through a heated glass cell containing
alkali vapor and the time and wavelength resolved fluorescence is detected
at a right angle. same wavelength asthe second exciting laser. For rapidly
decaying low lying states the fluorescence is typically stronger than the
scattered laser light, and for slowly decaying high lying states the phototube
can be gated on after the laser pulse. The longer lived Na ftp lifetimes can be
observed by exciting the ftd state which quickly decays, leaving an
observable fraction, �10%, of the initial ftd population in the much longer
lived ftp state.19 The time resolved 4p�>3s fluorescence can be directly
detected, but for higher ftp states the ftp �> 3s fluorescence has too short a
wavelength to pass through a glass cell. Instead, the time resolved fts^3p
fluorescence, the second step of the ftp -� fts -� 3p cascade is observed.
The long time behavior of the observed signal reflects the lifetime of the ftp
state, which is much longer than the lifetime of either the fts or ftd states.
Finally, the lifetimes of several Na ftf states have been measured using a
resonant microwave field to equalize the populations in the ftd and ftf
levels.20 Observing the time resolved fluorescence from either the ftf or ftd
state gives a decay rate indistinguishable from the ftd state decay rate,
indicating that the ftf lifetime isvery nearly the same asthe ftd lifetime. Had
theftflifetimes been much longer than the ftd lifetimes, this method would
have been relatively insensitive to the ftf lifetimes. The fluorescence methods
described above work well for states of principal quantum numbers less than
15. For higher principal quantum numbers they do not work so well for three
reasons. First, the amplitude of the fluorescence signal

Radiative lifetimes 47 uu 80 60 40 30 20 10 8 6 4 3 2 1 0.8 0.6 0.4 0.3 0.2 0

0.1 _ I O o o AA A � A � A B " � � + � � A � i A y � i i i i i i i i i i
i i i i i i i i i i i i i i 10 12 14 16 18 20 22 24 26 Fig.4.5 Naradiative lifetimes
vs n*. Experimental valuesfor ns(A), np(�),nd(�) andni (+) states are
shown. The lifetimes states below n = 15 have been measured by
fluorescence techniques, attemperatures ofapproximately 400 K. The
lifetimes of n > 15 states have been measured byfieldionization, the ns and nd
states at 30 K and the np states at 300K. The theoretical OK np lifetimes are
also shown (O). They are far above the measured valuesathighndue
toblackbodyradiation. Finally, thehydrogeniclifetimesare shown by theline
segments (�) (from refs. 4, 18-22, 30). decreases asn 6 , second, the atoms
pass out of thefieldof view due to their longer lifetimes, and, third, black
body radiation can appreciably alter the lifetimes.8 The time delayedfield
ionization method employed by Spencer etal.21 to measure the Na ns and nd
lifetimes circumvents all of these problems. They passed a well collimated
Na atomic beam through an interaction region cooled to 30 K. The atoms
were excited by two pulsed dye laser beams collinear with the atomic beam
and tuned to the 3s �> 3p and 3p�> ns or 3p�� nd transitions. At a
variable time after the laser excitation a field ionization pulse was applied to
atoms in the interaction region and the resulting ions detected. The atoms
which passed out of the interaction region between the laser pulse and the
field ionization pulse were

48 Rydberg atoms Table 4.1. Lifetime parametersfor alkali atoms. ua Na* Kb

Rbb Csb r0 (ns) a r0 (ns) a r0 (ns) a r0 (ns) a r0 (ns) a s 1.39 2.80 1.38 3.00
1.32 3.00 1.43 2.94 1.43 2.96 P 5.69 2.78 8.35 3.11 6.78 2.78 2.76 3.02 4.42
2.94 d 0.59 2.92 0.96 2.99 5.94 2.82 2.09 2.85 0.96 2.93 f 0.96 3.06 1.13
2.96 0.83 2.95 0.76 2.95 0.69 2.94 a (computed from values in ref. 29) *
(fromref.31) replaced by atoms which had been excited further upstream by
the laser. Using this approach they were able to make measurements at delay
times of up to 25 JUS, corresponding to about 2 cm of travel of the atoms. It
is useful to note that if the interaction region iscooled the only criterion for
thefieldionization pulse isthat it be strong enough to ionize the Rydberg state
in question. If significant black body redistribution occurs, as it does at
300K, thefield ionization pulse must be chosen with more care, as described
in Chapter 5. In Fig. 4.5 we show measured values of Na ns, np, nd, and ni
lifetimes vsn*3 on logarithmic scales. The scales differ by afactor of 3, so an
n*3dependence appears as a 45�line. As shown by Fig. 4.5 the n3
dependence suggested by Eq. (4.11) is obtained. The values for n < 15 have
all been obtained using fluorescence detection from atoms in a cell at about
425 K. For all but the np states the black body radiation isanegligible effect.
For the np stateswe also show the calculated 0 K values of Theodosiou, and
for the 7p state, the measured value is about 30% below the calculated 0 K
value and consistent with the calculated 425 K value of Theodosiou.29 For
n>15 all values have been obtained usingfieldionization. The ns and nd
values are those obtained by Spencer etal.21 at 30 K and are clearly in
reasonably good agreement with an extrapolation of the values obtained at
lower n using fluorescence detection. For the 18p and 19p lifetimes
measured at 300 K, however, black body radiation reduces the lifetimes far
below an n*3 extrapolation of the lifetimes of the lower np states.30 We also
show in Fig. 4.5 short 45�lines at n = 10corresponding to the lifetimes of
the hydrogenic ns, np, nd, and ni series. The Na np and nd lifetimes are
respectively factors of 50 and 2 longer than their hydrogenic counterparts,
while the Na nslifetimes are about 30% shorter than the H nslifetimes. The
Na and H ni lifetimes are the same, which is not surprising as the Na ni states
and the Na nd states to which they predominantly decay all have quantum
defects of nearly zero.

References 49 The lifetimes of alkali Rydberg states can be calculated

accurately, asshownby Theodosiou,29 who has made extensive comparisons
between observed lifetimes and calculated values. From this work is is also
apparent that the black body radiation correction of Eq. (4.16) is important
for high lying states. In light of these twopoints, it seems that the most
compact and consistent wayof presenting the alkali lifetimes is by afitof the
calculated 0 Kvalues to the form31 r = T0(�T -(4-17) The values of r0 and
a are presented in Table 4.1. As shown by Table 4.1, the values of a are all
near 3, as expected from Eq. (4.11). The biggest discrepancies occur in the
Li and K np series which have their Cooper minima nearest the ionization
limit. It is also apparent that the np lifetimes are the longest, as expected from
Figs. 4.1 and 4.2. In using the values of Table 4.1 it is worth remembering
that for high n states the black body decay rate is�independent, so the
lifetimes of the longest lived states are the most affected by black body
radiation.32 References 1. H. A. Bethe and E. A. Salpeter, Quantum
Mechanics of One and Two Electron Atoms (Academic Press, New York,
1957). 2. E. S. Chang, Phys. Rev. A 31,495 (1985). 3. U. Fano and J. W.
Cooper, Rev. Mod. Phys.40, 441 (1968). 4. A. Lindgard and S. E. Nielsen,
Atomic Data and Nuclear Data Tables 19, 533 (1977). 5. G. V. Marr and D.
M. Creek, Proc. Phys. Soc. (London) A 304, 233 (1968). 6. G. V. Marr and
D. M. Creek, Proc. Roy. Soc. (London) A 304, 233 (1968). 7. W. Sandner, T.
F. Gallagher, K. A. Safinya, and F. Gounand, Phys.Rev. A 23, 2732 (1981).
8. M. J. Seaton, Proc. Roy. Soc. (London) A 208, 408 (1951). 9. G. Baum,
M. S. Lubell, and W. Raith, Phys.Rev. A 5,1073(1972). 10. R. J. Exton, J.
Quant, Spectrosc. Radiat. Transfer 16, 309 (1975). 11. G. Pichler, J. Quant,
Spectrosc. Radiat. Transfer 16, 147 (1975). 12. C. J. Lorenzen and K.
Niemax, /. Phys. B 11,L723 (1978). 13. J. M. Raimond, M. Gross, C. Fabre,
S. Haroche and H. H. Stroke, /. Phys.B 11,L765(1978). 14. D. W. Norcross,
Phys. Rev. A 20, 1285 (1979). 15. C. M. Huang and C.W. Wang, Phys.Rev.
Lett. 46, 1195 (1981). 16. W. Hansen, /. Phys. B. 16, 933 (1983). 17. D.
Kaiser, Phys.Lett. 51A, 375(1975). 18. T. F. Gallagher, S. A. Edelstein and
R. M. Hill, Phys. Rev. A 11, 1504 (1975). 19. T. F. Gallagher, S. A.
Edelstein and R. M. Hill, Phys.Rev.A 14, 2360(1976). 20. T. F. Gallagher,
W. E. Cooke, and S. A. Edelstein, Phys.Rev. A 17, 904 (1978). 21. W. P.
Spencer, A. G. Vaidyanathan, D. Kleppner, and T. W. Ducas, Phys.Rev. A 24,
2513 (1981). 22. T. F. Gallagher and W. E. Cooke, Phys. Rev. A 20, 670
(1980). 23. F. Gounand, P. R. Fournier, J. Cuvellier, and J. Berlande, Phys.
Lett.59A, 23 (1976). 24. F. Gounand, M. Hugon, and P. R. Fournier, /. Phys.
(Paris) 41, 119 (1980). 25. M. Hugon, F. Gounand, and P. R. Fournier, /.
Phys. B 11, L605(1978). 26. H. Lundberg and S. Svanberg, Phys. Lett. 56A,
31 (1976). 27. K. Marek and K. Niemax, /. Phys. B 9, L483 (1976). 28. J. S.
Deech, R. Luypaert, L. R. Pendrill, and G. S. Series, J. Phys.B 10,
L137(1977). 29. C. E. Theodosiou, Phys. Rev. A 30, 2881 (1984). 30. T. F.
Gallagher and W. E. Cooke, Phys.Rev. Lett. 42, 835 (1979). 31. F. Gounand,
/. Phys. (Paris) 40, 457(1979). 32. W. E. Cooke and T. F. Gallagher,
Phys.Rev. A 21,588 (1980).

5 Black body radiation Lying >4 eVabove the ground state, Rydberg states
are not populated thermally, except atveryhightemperatures. Accordingly,
itisnatural toassumethat thermal effects are negligible in dealing with
Rydberg atoms. However, Rydberg atoms are strongly affected by black body
radiation, even at room temperature. The dramatic effect of thermal radiation
isdue to two facts. First, the energy spacings AW between Rydberg levels are
small, so that AW < kT at 300 K. Second, the dipole matrix elements of
transitions between Rydberg states are large, providing excellent coupling of
the atoms to the thermal radiation. The result of the strong coupling between
Rydberg atoms and the thermal radiation is that population initially put into
one state, by laser excitation for example, rapidly diffuses to other
energetically nearby states by black body radiation induced dipole
transitions.1"3 Both the redistribution of population and the implicit increase
in the radiative decay rates are readily observed. Although the above
mentioned effects on level populations are the most obvious effects, the fact
that a Rydberg atom is immersed in the thermal radiation field increases its
energy by a small amount, 2 kHz at 300 K. While the radiation intensity is
vastly different inthe two cases, this effect is the same as the ponderomotive
shift of the ionization limit in high intensity laser experiments. Black body
radiation The most familiar way of representing black body radiation is to
use the Planck radiation law for the energy density, or the square of the
electric field, p(v). Explicitly4 8jthv3 p{v)dv = c\e hv/kT -iy ( } where k is
theBoltzmann constant, his Planck's constant, visthefrequency of the black
body radiation, and T is the temperature. In Fig. 5.1we showp(v)vsv at 300
Kusing both frequency and wavenumber as abscissae. A typical optical
transition from the ground state of an atom has frequency v = 3 x 1014 Hz,
and a transition between two Rydberg states has frequency v ~ 3 x 1011Hz.
Thus, it isapparent from Fig. 5.1 that to aground state atom the black body
radiation appears as a slowly varying, nearly static field, whereas to a
Rydberg atom it appears to be a rapidly varying field. 50

Black body radiation 51 X"1 (103 cm"1) 0.5 1.0 1.5 Fig.5.1Energy density
p(v)as a function of frequency vat 300 K (from ref. 5). While the
representation of black body radiation given inEq. (5.1) and Fig. 5.1 isthe
most familiar, it is not the most useful for the most important effect of black
body radiation on Rydberg atoms, inducing transitions between neighboring
levels. To calculate the rates of these transitions it is more convenient to
express the radiationfieldin terms of the number of photons per mode of the
radiationfield, i.e. the photon occupation number n of each mode. The photon
occupation number n is given by4 At low frequencies, for which hv � kT,
Eq. (5.2) reduces to hv Fig. 5.2 showsthe dependence of n on vfor T =300 K,
again with both frequency (Hz) and wavenumber (cm"1) used asabscissae. It
is useful torecallthat at 300 K, kTlh ^ 6 x 1012Hz or kTlhc � 200 cm"1. This
point isapparent in Fig. 5.2, since n = 1 at X~x = 200 cm"1. Since the
vacuum fluctuations, which lead to spontaneous emission, are given by n =
1/2, black body radiation at frequencies greater than kT/hn, where n � 1,
does not lead to significant effects. For an atom in its ground state with
transitions at ~ 104 cm"1, black body induced transitions are unimportant,
since n � 1. However, for a Rydberg state with transitions at 10 cm"1,
where n ~ 10,the black body induced transition rates can

52 Rydberg atoms Fig. 5.2 Occupation number nas afunction of frequency v

at 300 K(from ref. 5). be an order of magnitude larger than the spontaneous
emission rates, and the effect of black body radiation is significant. Eqs.
(5.1)-(5.3) present black body radiation in a familiar form. Both to conform
to the general use of atomic units in this book and to simplify the calculation
of the Rydberg atoms' response to the radiation we reexpress Eqs.
(5.1)�(5.3) in atomic units.5 We reexpress Eq. (5.1) as , XJ 2a3a)3da) ( ) d
(54 > where a is the fine structure constant and co is the energy in au.
Similarly, we may reexpress Eq. (5.2) as and Eq. (5.3) for w�kT as - kT �
- - � (5.6) We now consider the transition rates for absorption of and
stimulated emission induced by the black body radiation and compare these
rates to the spontaneous

Black body induced transitions 53 emission rates. This comparison provides

a reasonable indication of the importance of the effects of black body
radiation induced transitions for Rydberg atoms.6 Black body induced
transitions Asshowninthepreviouschapter, the spontaneous decayrate ofthent
statetothe lower lying n'�'state is given by the Einstein A coefficient An>v
n�. 7 In the presence of thermal radiation the stimulated emission rate
Kn^^n�, is simply n times aslarge asthe spontaneous rate. Explicitly,
Kn'�',n� = ^An'�',n^ (5-7) where n is implicitly evaluated at the frequency
conf� r ,n�' It is convenient to reexpress Eq. (5.7) in terms of the average
oscillator strength, where a is the fine structure constant, o)n>e>nt is the
energy difference Wn.e. - Wnt, and where �max is thelarger of �and�'.
The black body photons can also be absorbed asthe atoms in the nt state make
the transition to a higher lying n't' state. Both the stimulated emission and
absorption rates are givenby Because of the variation of n with frequency, the
frequency dependence of Kn>t',nt is quite different from that of An^^n�, and
these two processes favor different final states. This point ismade
graphically inFig. 5.3, which is aplot of An>PilSsand Kn>PflSsvsn' forthe
Na 18s state and T= 300 K. As shown inFig. 5.3, black body radiation favors
transitions to nearby states and spontaneous emission favors transitions to the
lowest lying states. In addition to driving transitions todiscrete states the
black body radiation can also photoionize the Rydberg atoms. The
photoionization rate, 1/T��isgiven by8 The matrix elements aretheradial
matrix elements of rbetween theinitial nt state and continuum �� 1 and�+
1 waves which arenormalized per unit energy. Just aswepreviously summed
the spontaneous emission rates over the possible final states toobtain thetotal
spontaneous decay rate l/rw�, we can sum the black

54 Rydberg atoms 2 4 6 8 10 12 14 16 18 20 22 Fig. 5.3Spontaneous

emission rates of the 18s state tolower lying n'p states,An^p 18s, (D) and the
300K black body transition rates of the 18sstate to higher and lower lying n'p
states, 7�n'p 18s, (�) asafunction of n' (from ref. 5). body radiation
transition rates to compute the total black body radiation induced decay rate
1/r^. Explicitly, -4 = 2a3 V no?n.VM \fn,e,M\ + -�r, (5.12) and the total
decay rate l/rj� is given by ^=r+^" (5 -13) From Fig. 5.3 it is apparent that
the most important contributions to the black body radiation decay rate are
from transitions to nearby states for which \con'^^\ < kT. In this case we may
substitute Eq. (5.6) for n and write l/r)$as -4 = 2a3kTjjWn^nJn^. (5.14) In
Eq. (5.14) the summation over n' implicitly includes the continuum. For a
one-electron atom we may use the sum rule7 2_ 3n2 which we may use to
write 1 Aa3kT X 2 Wn'e'Mfn't'M = �2, (5.15) 3n rS 3n2 (5-16) For T = 300
K, Eq. (5.16) is accurate to 30% for n > 15, and asn isincreased Eq. (5.16)
becomes an increasingly good approximation. More important, it brings
Black body energy shifts 55 K 3 I I I I I � _ I 16 18 20 22 n' Fig. 5.4
Spontaneous emission rate of the n = 18, � =17state to then = 17,i = 16state,
A17�=16 i8^=i7, (�) and the 300Kblack body transition rates ofthe n = 18,
�= 17stateto '�'�'states of �'= � � 1,/�n'^5ig�=i7 >(*) as a function
of n' (from ref.5). out two interesting points. First, the total black body decay
rate Vrnebb does not depend on �, only on n. Since the radiative decay rate
decreases rapidly as � is increased,9 for high �states the black body
induced decayrate is often much larger than the spontaneous emission rate.
This point is shown graphically in Fig. 5.4 in which the spontaneous
emission rates and black body radiation emission and absorption rate are
shown for the n = 18,� =17 state. In contrast to the 18s state, for which l/rbb
~ 0.2/r, the ft = 18,� =17 state has ablack body decay rate which is ten times
larger than its spontaneous decay rate. Second, because of the IIn2 scaling of
the black body radiation decay rate, it is evident that even for low � states
that at high enough n the black body radiation rate exceeds the spontaneous
emission rate, which scales as Vn3. Even at n =20,the black body radiation
rate is at least 20% of the spontaneous emission rate for all � states, and in
many experiments it is a significant source of population redistribution.
Black body energy shifts That part of the black body spectrum coincident
with the atomic transition frequencies leads to the transitions which
redistribute the population. In contrast, all the energy of the black body
radiation contributes to the shift of the energy levels. The energy shift is a
second order ac Stark shift, and for state n the shift , is given by10 e'-vl)

56 Rydberg atoms where EWh is the electric field of the black body
radiation in a bandwidth dcoh at frequency cob. Since E2OJb is proportional
to the energy density, examining Fig. 5.2 we can see that at room temperature
the peak of the black body radiation energy density is at ~500 cm"1, far
above the strong n ~ 20 transitions which have wavenumbers of �20 cm"1.
In fact, for atoms in Rydberg states it is often true that the frequencies of the
strong transitions are much lower than the frequency of the black body
radiation, i.e. o)n^n^> � &>b, in which case we can ignore a)nz,n>z> in the
denominator of Eq. (5.17) and rewrite AWn�as AW^2>>�](| ^ p j - (5.18)
Using the oscillator strength sum rule7 7*'O,�=l (5.19) n'i' and integrating
over a>h in Eq. (5.18), we may express Eq. (5.18) as �Wn� = ^a\kT)2.
(5.20) Eq. (5.20) is accurate to -10% for n > 15 at T = 300 K. Eq. (5.20)
indicates that all Rydberg states experience the same energy or frequency
shift, which is +2.2 kHz at 300 K. Eq. (5.20) corresponds to the energy shift
of a free electron in an oscillating electric field. This point ismore apparent
if we simply use the oscillator strength sum of Eq. (5.19) and apply it to the
case of a monochromatic field of angular frequency coh for which JQ
Eb2dcob = E2. In this case Eq. (5.18) can be written (521) which is the
average kinetic energy of a free electron in a field E cos cot. The equality of
the energy shift of a free electron and an electron in a Rydberg state is not
surprising since the physical effect of the ac black body field, at a frequency
high compared to the orbital frequency of the electron, is to superimpose a
fast, ~&>b, wiggle on the orbital motion of the electron. The energy of the
wiggle motion is independent of the much slower orbital velocity. While the
black body radiation energy is mostly at frequencies high compared to the
Rydberg state frequencies, it is low compared to the frequencies of
transitions of low lying states. Explicitly, &Vr,n�>>a)b, and for low lying
states we can ignore cob in the denominator of Eq. (5.18). In this case the
Stark shift is equal to that produced by a static field and can be expressed as
AWn( = y _L^nt r E^do^ fe J 4

Initial verification 57 For a one-electron Rydberg atom we may use the

hydrogenic sum rule11 'n^nLj (5.23) so that we may write Eq. (5.22) as AWn
= - f9- E2abAwb = - \ {an)\kT)\ (5.24) JO J Evaluation of Eq. (5.24) gives
ashift of -0.036 Hz at 300 K, whichis negligible for all practical purposes.
Finally, we must remember that for some states it will be true that avr ,nt~ w
b> in which case neither of the above approximations isvalid and Eq. (5.17)
must be evaluated explicitly ashas been done byFarley and Wing.12At 300
Kcon^^n+1^ ~ cob for n ~ 8. If we confine our attention to Rydberg states of
n > 15,for which Eq. (5.20) is valid, it is apparent that all the Rydberg states
and the ionization limit are shifted up in energy by an amount which
isproportional to the square of the temperature. If the atoms are exposed to a
monochromatic radiation field satisfying the same frequency criteria, the
Rydberg state energies and the ionization limit are shifted according to Eq.
(5.21), which is often termed the ponderomotive energy shift.13 In this
chapter we have implicitly assumed the Rydberg atom to be a one electron-
atom. In the perturbed Rydberg series of, for example, alkaline earth atoms,
Rydberg states can have mixed valence-Rydberg character. In such states the
black body effects are reduced by a factor equal to the fractional Rydberg
character.14 Initial verification The experimental observations which called
attention to the effects of room temperature black body radiation on Rydberg
atoms were made using Rydberg states of Xe and Na.1"3 However, Pimbert
had earlier noted the effect of higher temperatures.15 Two kinds of
experiments were performed, measurements of population redistribution and
increased decay rates. Both of the effects measured could conceivably arise
from collisions, and this possibility was systematically ruled out in the
experiments. In the Xe experiment atoms in a metastable beam were excited
to the 26f state by a pulsed laser and exposed to the thermal radiation for
periods of 1.5, 7.5, and 15.5 JUS, after which a ramped field was applied to
the atoms. This procedure leads to the field ionization signals shown in Fig.
5.5.l For each time delay there is a clear progression of peaks in
thefieldionization spectra which are assigned the nvalues shown. At longer
delay times the intensities of the

58 Rydberg atoms mi11 i i�i r t=l.5/xs in 11 i i i i t= 15.5/is 35 30 28 27 26

PRINCIPAL QUANTUM NO. , , , ! , , i i 1 500 1000 FIELD
STRENGTH,V/cm Fig.5.5Field ionization spectra atdifferent times after
thelaserpulse excites Xe atoms to the 26f level showing the black body
radiation induced transfer to higher levels (from ref. 1). n > 26 peaks grow,
aswould be expected for progressively longer exposure times to the thermal
radiation. In addition, due to the fact that the black body radiation induced
transitions are dipole transitions, only d and gfinalstates are expected. The
fact that the field ionization spectra exhibit a sharp peak for each nis
consistent with only one or two, not all, �states being populated by the
thermal radiation.1 If allthe ( stateswerepopulated there would be no sharp
peaks inFig. 5.5. In an analogous experiment inwhich the initial population
was placed in theNa 18s state,the transferred population wasfound tobe
inthehigher lyingp states,as expected for dipole transitions. Furthermore, the
fraction of the atomswhich had undergone transitions to each of the higher p
states after 5 jus was in good agreement with values calculated on the basis
of 300 K black body radiation induced transitions, as shown by Table 5.1.2
The other initial experiment was the observation ofthe increase in the decay
rate due to 300 K black body radiation, shown by Eq. (5.13).3The decay rate
of
Temperature dependent measurements 59 Table 5.1. Calculated and observed
populations in higher lying np states 5JUS after the initial population of the
Na 18s stateexpressed in terms of the initial 18s population." Final state 18p
19p, 20p >20p,ep Total Calculated yield 4.2% 1.6% 1.0% 6.8% Observed
yield 5.0% 1.6% 1.2% 7.8% a (see ref. 2) Table 5.2. Lifetimes of the Na Up
and 18p states.61 State 17p 18p "(from *(from ref. ref. r (jus) 48.4* 58.4* 3)
17) rbb 22.7 25.6 rT (calc) 15.5 17.9 rT(obs) 13.9+1;| the Na 18s state at
300Kdiffers from its 0 Krate by about 20%, which is not a large enough
difference that a measurement at 300 K alone is entirely convincing. The Na
np states of n ~ 20, on the other hand, have long lifetimes1617 and offer a
reasonable prospect for a convincing measurement, even if made only at 300
K. The Na 17pand 18plifetimes were measured by exciting Na atoms in
abeam to the 17p and 18p states and observing the populations in these states
as a function of time after the laser pulse. The results of the experiment are
given in Table 5.2, which shows clearly that the experimentally observed
values are in good agreement with the lifetimes calculated with the black
body radiation, but in stark disagreement with the 0 K lifetimes. Temperature
dependent measurements Subsequent to the initial 300 K observations,
measurements were done as a function of temperature. The first
measurements, by Koch et a/.,18 were done at

60 Rydberg atoms elevated temperatures, and later measurements were done

at temperatures below room temperature. A critical issue in such
measurements isknowing the temperature the atoms are experiencing.
Typically an enclosure cooled to a temperature from 6 K to 300 K surrounds
the atoms, shielding them from the 300 K radiation from the walls of the
vacuum chamber. However, simply making an enclosure in which 90% of the
solid angle seen by the atoms is cold isnot adequate, because at far infrared
wavelengths many materials, and virtually all metal surfaces, are excellent
reflectors. Thus a 10% aperture to the warm world outside is adequate to
raise substantially the temperature the atoms in the enclosure experience.
Two approaches have been used to ensure that the temperature experienced
by the atoms isclose to the temperature of the enclosure. Hildebrandt etal.
lined the inside of the enclosure with graphite coated Cu wool, which they
verified separately to be an effective way of absorbing far infrared
radiation.19 An alternative approach was followed by Spencer et al.20They
covered the apertures through which laser beams propagated with glass,
which is opaque at far infrared wavelengths, and those through which the
atomic beam propagated with a fine, 70jum, mesh, which blocks all radiation
with X > 140 jum. Experiments with Xe analogous to the previously
described Xe experiment were carried out at both 90 K and 300K.
Specifically the time dependences of the populations in higher nd and ng
states were measured subsequent to population of an ni state by a pulsed
laser.19 The time dependences of the observed populations were fit to a
model which yielded the radiative transfer rates from the initially excited
state to the final states. Not surprisingly, as the temperature was reduced
from 300 K to 90 K, not only was the overall radiative transfer reduced, but
the transfer to the highest lying states wasmost sharply reduced, asexpected
from the dependence of the photon occupation number onfrequency and
temperature. For example the transfer from the Xe 25f to 26d,g states was
reduced by a factor of 3 while the 25f-27d,g transfer was reduced by a factor
closer to 5. Spencer et al. populated the Na 19s state and observed the sum
of the populations in the 19s, 19p, and 18p states, at times from 2-32 jus after
the laser excitation, to determine the sum of the radiative transition rates from
the 19s to 18p and 19p states.20 The measurements were made at
temperatures from 6 K to 210 K with the results shown in Fig. 5.6. As shown
by Fig. 5.6, the observed rates agree well with the calculated rates, shown by
the line. At 0 K the rate is entirely due to the spontaneous 19s �> 18p
transition, and the increase above this rate is due to thermal radiative transfer
to both the 18p and 19pstates. It isinteresting to note that above ~ 30Kthe
transfer rate increases linearly with T asexpected from Eq. (5.16). The linear
behavior only occurs above 30 K since only then iskT > co for these two
transitions. While the Na 19s �> 18p, 19p transfer rate is an excellent
illustration of the validity of Eq. (5.16) for co < kT, the photoionization of
Rydberg atoms of n ~ 20 by black body radiation is a test of the regime co ~
kT, and this regime has been explored by Spencer et al.21 Specifically, they
measured the relative rates for

Suppression andenhancement of transition rates 61 50 100 150 200 250

TEMPERATURE (K) Fig. 5.6 Experimental and theoretical transfer rate of
the Na 19s state to the 18p and 19p states vs temperature. The solid line is
calculated with no adjustable parameters (from ref. 20). photoionization of
the Na 17d state by thermal radiation at temperatures of 90-300 K. The
experiments were done by exciting the Na 17d state, allowing
photoionization to occur for 500 ns, a time ten times shorter than the 17d
lifetime, and then collecting the photoions by applying a smallfieldpulseof 8
V/cm.A field this small was verified to produce negligible field ionization of
bound states excited by black body radiation. When the relative rates are
normalized to the calculated rate at 300 K, they are found to be in excellent
agreement with the calculated rates. Since the 17d state is bound by 380
cm"1, which exceeds kT, the photoionization rate varies rapidly with
temperature, roughly a factor of 100 between 90 K and 300 K, as shown
byFig. 5.7. Therapid, nearly exponential, dependence of the photoionization
rate on temperature shown inFig. 5.7 is to be contrasted with the linear
temperature dependence of the 19s �� 19p, 18p transfer rate shown in Fig.
5.6. Suppression and enhancement of transition rates We have described the
effects of black body radiation in free space. In a closed cavity the radiation
is confined to the allowed modes of the cavity. In essence all the thermal
radiation is forced into the cavity modes, raising the intensity at the

62 Rydberg atoms 1000^ � 100150 200 250 300 TEMPERATURE (K) Fig.
5.7 Black body induced photoionization rate vs temperature for the 17d state
in Na. Scalesarelogarithmic. The solidlinerepresents the calculated
values.Experimental points are normalized tothecalculated value at
300K(from ref. 21). frequencies of the modes by afactor of Qrelative to the
free spacevalue.22 Here Q is the quality factor of the cavity, the frequency of
a mode divided by its linewidth (FWHM), i.e. Q = v/Av. Similarly, between
the cavity modes the intensity is substantially depressed. At a cavity mode
resonance the radiation transfer rate is increased, by a factor of <2, and away
from a resonance, the transfer rate is suppressed. As pointed out by
Kleppner, Rydberg atoms, with their long wavelength transitions, provide a
nearly ideal system for the study of such notions.23 The first experimental
realization of this notion wasbyVaidyanathan etal. who excited Na atoms to
nd states between two parallel plates 0.337 cm apart.24 The region between
these plates is cutoff for propagation of radiation of wavelengths longer than
0.674 cm polarized parallel to the plate surfaces, and such radiation is not
present between the plates. Stated another way, frequencies less than the
cutoff frequency vc = 1.48 cm"1 do not propagate. Radiation polarized
perpendicular to the surfaces can propagate freely between the plates. The
zero static field 29d-30p interval isjust under 1.48 cm"1. However, the
application of a small staticfieldincreases the separation of the levels, and
aspacing of 1.48 cm"1 occurs at astaticfieldof 2.4 V/cm. Thus asthe
staticfieldisincreased from zero to beyond 2.4 V/cm the 29d �� 30p
transition frequency passes from below to above cutoff, and the black body
radiation transition rate increases accordingly. In the experiment Na atoms
are excited by two lasers from the ground state to the nd state in an
environment cooled to 180 K at which point n = 86 for the 29d�> 30p
transition. The atoms are exposed to the thermal radiation for 20^s,

Suppression and enhancement of transition rates 63 2 3 4 FIELD (V/cm) 108

1.10 1.12 114 1.16 118 FREQUENCY (v/vc) FREQUENCY!*/^) Fig. 5.8
Black body radiative transfer signals in Na located between parallel
conducting plates for 29d-^ 30p (left-hand side) and28d-� 29p (right-hand
side) asafunction of the absorption frequency. Thecutoff frequency isvc =
l/2d =1.48 cm"1, where d is the plate separation. The increase in the transfer
rate at v = vc (left-hand side) is due to the "switching on" of the radiation
polarized parallel totheplates (from ref. 24). after which they are exposed to
a field ionization pulse to ionize atoms in the (n + l)p state selectively. In Fig.
5.8 we show the results obtained when the 28d and 29d states are excited as
functions of static field over the range 0-5 V/cm. As shown in Fig. 5.8, the
29d-30p interval is tuned through the cutoff frequency of the plates in this
field range. Correspondingly, there is a sharp increase in the population
found in the 30p state when the field tunes the 29d-30p interval through the
cutoff frequency. In contrast, the 28d-29p interval always exceeds the cutoff
frequency, and no sharp increase in the 29p population is observed as the
field is tuned. Just asthe black body transitions can be suppressed, they can
alsobe enhanced, by tuning to a cavity resonance. Raimond et al.25 have
passed a thermal beam of Na atoms excited to the 30s1/2state through a300
KFabry-Perot cavity, which can be tuned to the 30s1/2-30p1/2 resonance at
134 GHz, and detected the atoms downstream by field ionization. With the
cavity tuned off resonance 5% of the atoms make the 30s1/2 �� 30p1/2
transition, but when it is tuned to resonance an additional 5% make the
transition, in agreement with the calculated value.25 In any low angular
momentum state the radiative decay rate is usually dominated by the high
frequency transitions to low lying states, and as aresult it is impossible to
control completely the decay rate using a millimeter wave cavity. In a
circular � = m = n - 1state the only decay is the far infrared transition to the
n � 1 level, and Hulet et al. have observed the suppression of the decay of
this level.26They produced abeam of Cs atoms in the circular n = 22,� = m
=21 state by pulsed laser excitation and an adiabatic rapid passage
technique.27 The beam of circular state atoms then passed between a pair of
plates 6.4 cm wide, 12.7 cm long, spaced by 230.1 /urn, and held at 6K. The
0Kradiative lifetime is 460^s, and

64 Rydberg atoms the transit time of the atoms through the plates to the
detector is approximately one lifetime at the peak of the velocity distribution.
In zero static field the wavelength Xof the n = 22, � = 21 -� n = 21, � = 20
transition is 450/rni, andA = 0.98(2d) where dis theplate spacing. However,
using the plates to apply a static field it is possible to reduce continuously the
energy separation between the levels by up to 4%, increasing the wavelength
of the transition to 1.02(2d). In other words it is possible to increase the
wavelength of the transition from below cutoff to above cutoff byvarying the
tuning field. At this point it is worth recalling that in the static field the �
=21, m = 21circular n = 22 state must interact with radiation polarized
parallel to the field plates to decay to the n = 21,� = m = 20 state. The
radiation polarized perpendicular to the plates cannot drive this transition.
When the wavelength is longer than the cutoff wavelength ,i.e. X > 2d,
radiative decay issuppressed, i.e. rcav = ��, but when the wavelength is
less than the cutoff wavelength X < 2d the radiative decay rate is increased
by 50%.26 They recorded time offlightspectra of atoms remaining in the
circular state asa function of the spacing of the plates and the tuning field
applied, and from these spectra determined the decay rates. Decay rates were
measured and calculated for three conditions. First the plate spacing was
increased by a factor of 30 to observe the free space decay. They observed a
lifetime of 450(10) jus, in good agreement with the calculated 6 K value of
451 jus. With the plate spacing returned to 230.1 jum the decay rates were
measured with the electricfieldset so asto tune the wavelength both below
and above the cutoff wavelength. With alowelectricfield, for which X< 2d,
the measured decay rate was 50% larger than the free space value, to within
5%. When the field was increased so that X > 2d, roughly twice the number
of atoms was detected and the decay rate was consistent with zero, to within
5% of the free space decay rate. Level shifts Rydberg states which are pure
Rydberg states, not containing admixtures of valence states, all have the same
positive energy shift in a thermal radiation field. To measure this shift
requires careful measurement of a transition from a long lived low lying
atomic state. Holberg and Hall have measured the thermal shift of Rb atoms
exposed to a thermal source which ranged in temperature from 350 K to 1000
K using two photon, Doppler free spectroscopy between the ground state and
the Rydberg states.28 In their experiment a Rb beam passes through the two
arms of a folded optical cavity, so the atoms interact with the optical
radiation twice, providing a Ramsey interference pattern. An attraction of the
Ramsey method is that the Ramsey interference pattern is located at the
optical field free location of the 5s �� ns transition. It is thus unaffected by
laser

Experimental manifestations and uses 65 light shifts. The central Ramsey

interference fringe is 40 kHz wide, the inverse of the transit time of the atoms
between the two arms of the optical cavity. The radiation from the heated
black body source is focused on the atomic beam between its intersections
with the two laser beams. Due to the small solid.angle subtended by the
heated black body source at the intersection of the atomic and laser beams, it
is only �10% as effective as a complete enclosure at the same temperature.
The radiation from the heated black body source is chopped and the position
of the central fringe of the Ramsey pattern observed with and without the
thermal radiation. In spite of the reduced efficiency of the source in
producing a shift, blue shifts of up to 1.4 kHz were observed, at 876 K. The
blue shift increases with temperature as predicted by Eq. (5.20), and while
the results do not provide a stringent test, they are clearly consistent with Eq.
(5.20). The black body shifts are not confined to Rydberg atoms, but also
alter the frequency of atomic clock transitions. Itano etal.have shown that the
Cs 9 GHz ground state hyperfine interval, the definition of the second, is
increased by one part in 1014when the temperature is raised from 0 Kto 300
K.29. Experimental manifestations and uses As we have already pointed out,
the presence of black body radiation invariably decreases the lifetime of a
Rydberg state, and the decreased lifetimes have been reported in many
systems.3'30"32 What is most important, though, is the transfer of population
to nearby states, and this effect must be taken into account in making
measurements. To illustrate this point we consider the measurement of the
300 K lifetime of a Rydberg state using selective field ionization, an
apparently straightforward measurement.6 The 18s state of Na is calculated
to have a 0 K radiative lifetime of 6.37 JUS, which is reduced to 4.87JUS by
the 300 K black body radiation. Na atoms were excited to the 18s state at
time t = 0, and the population remaining at later times was determined by
applying an electric field pulse high enough to ionize easily the atoms in the
18s state. This measurement yielded an apparent lifetime of 7.8 JUS, a value
longer than the 0 K radiative lifetime. The problem was that the field also
ionized all the long lived, np states of n > 17 to which population was
transferred by black body radiation. Only when the signal from higher lying
states wasdetected separately was it possible to correct the observed 18s
signal for the presence of the long lived np states. The final value was
4.78jus, in excellent, and probably fortuitous, agreement with the calculated
value of 4.87 JUS. As shown by Table 5.2, at 300 Kthe n = 17and n = 18
black body decay rates are ~5 X 104 s"1, which might seem tolerable.
However, in even slightly dense

66 Rydberg atoms samples thermal radiation can trigger superradiance,

which depopulates an initially populated state in 10 ns.33'34 For
superradiance to occur between two levels (crudely speaking, laser action
without mirrors), a necessary condition is that the gain, G, be 1 along the
length of the sample. This criterion may be expressed as33 G = NLo>l,
(5.25) where N is the difference inpopulation densities of the upper and
lower levels, L is the length of the sample, and cr, the optical cross section,
is given by35 where g�and gu are the degeneracies of the upper and lower
states. Au� is the Einstein coefficient for the transition, and A is the
linewidth of the transition, to which the main contributions are usually
radiative and Doppler broadening. Because of the long wavelength of the
transitions between Rydberg states, the above gain criterion for
superradiance can be met with a very low density of atoms, hundreds of
atomsin a volume of 10~3 cm"1, and is easilymet even in a low density
atomic beam. The mere presence of adequate gain does not, however, ensure
superradiance. Some initial photons are required to trigger the
superradiance, and due to the low spontaneous transition rates of the long
wavelength transitions between Rydberg states, they are unlikely to come
from spontaneous emission. On the other hand, the omnipresent black body
radiation easily triggers superradiance. Gounand et al. have used
superradiant cascades initiated by black body radiation to populate
efficiently states which would normally be inaccessible using optical
excitation.34 In vapor cell experiments they observed the population in Rb
Rydberg states to undergo several very rapid superradiant cascades to
populate lower lying levels. The population istransferred in times far shorter
than would be possible by simple radiative decay.34 Fig. 5.9 shows the
population in various Rb Rydberg states after the initial population of the 12s
state. Normal radiative decay would require ~3/^s to populate the 7f state,
but as shown in Fig. 5.9, only 20 ns is required when superradiant cascades
occur. Stoneman and Gallagher have used black body radiation to make
precise measurements of the avoided crossingsbetween the K nsand n � 2
Stark manifold states in electric fields,36 and these measurements are
described in Chapter 6. Far infrared detection An interesting potential
application of Rydberg atoms is as a far infrared (or microwave) detector,
anotionfirstsuggested by Kleppner and Ducas.37 The basic

Far infrared detection 67 Fig. 5.9 Time dependence of the population in Rb

Rydberg states after the initial population of the 12sstate showing the rapid
superradiant cascades (from ref. 35). idea of the most straightforward type of
device is to use alaser to make atargetof Rydberg state atoms in one state, A.
This target is exposed to the radiation to be detected at the infrared frequency
coAB, which is equal to the frequency of the atomictransition from A to a
higher energy state B. Ifthe density of atoms in state A in the target is high
enough that the target is optically thick to the radiation at frequency coAB, a
photon at this frequency will be absorbed with nearly unit probability so that
one atom will undergo the transition to state B. The fact that the atom has
undergone the transition can be detected in either of twoways. The first is
selectivefieldionization of atoms in state Bbut not those instateA, and the
second is to detect the opticalfluorescencefrom state B but not from stateA.38
In either case each incoming photon is converted into an ion and an electron
or a visible photon, all of which are easily detected. Several variants of this
approach have been examined.2'39"41 With optical detection it is
unimportant whether state B lies above or below stated. If BliesbelowA,
asecond approach, triggered superradiance, ispossible, and the technique has
been carefully examined by Goy etal.42 To detect the same radiation at coAB
requires an optically thick target of atoms in the upper state A, which in a
cooled environment would not become superradiant, since spontaneous
emission is too slow toinitiate theprocess. However, anincoming photon at
frequency coAB would trigger a superradiant avalanche, converting one
incoming photon into ~104 outgoing photons and leaving 104 atoms in state
B, which would be trivial to detect. As shown by the sensitivity given in
Table 5.3, this approach is equally as sensitive as the more straightforward
approach described above. While Rydberg atoms have been used to detect
infrared radiation with wavelengths as short as 2.34 jum, they are most
effective as detectors of far infrared

) Rydberg atoms Table 5.3. Far infrared detection sensitivity in Rydberg

atom experiments. Method Absorption,0 field ionization Absorption,b field
ionization Absorption,0 field ionization Triggered^ superradiance "(seeref.
2) 6 (seeref. 41) c (seeref. 40) ^(seeref. 42) Source 300 K black body Far
infrared laser Variable temperature black body Klystron harmonic
Wavenumber cm"1 22 20 3.3 3.6 NEP W/cm2Hz1/2 1Q-14 5 x KT15 lO"17
3 x 10~17 radiation, and in Table 5.3wepresent measured sensitivities for
Rydberg atom detection of far infrared radiation. References 1. E. J. Beiting,
G. F.Hildebrandt, F.G. Kellert, G. W. Foltz, K. A. Smith, F. B.Dunning, and
R. F. Stebbings, /. Chem. Phys.70, 3551 (1979). 2. T. F. Gallagher and W. E.
Cooke, Appl. Phys. Lett. 34, 369 (1979). 3. T. F. Gallagher and W. E. Cooke,
Phys. Rev. Lett. 42, 835 (1979). 4. R. Loudon, TheQuantum Theory of Light
(Oxford University Press, London, 1973). 5. T. F. Gallagher, in Rydberg
States ofAtoms and Molecules, eds. R. F. Stebbings and F. B. Dunning
(Cambridge University Press, Cambridge, 1983). 6. W.E. Cooke and T. F.
Gallagher, Phys. Rev. A 21, 588 (1980). 7. H. A. Bethe and E. A. Salpeter,
Quantum Mechanics of One-and-Two-Electron Atoms (Academic Press,
New York, 1957). 8. U. Fano and A. R. P.Rau, Atomic Collisions and
Spectra (Academic Press, New York, 1986). 9. A. Lindgard, andS.
A.Nielsen, Atomic Data and Nucl.Data Tables 19,534 (1977). 10.
C.H.Townes andA. L. Schawlow, Microwave Spectroscopy (McGraw-Hill,
New York, 1955). 11. U. Fano and J. W.Cooper, Rev. Mod. Phys.40, 441
(1968). 12. J. W.Farley and W. H. Wing, Phys.Rev. A 23, 2397(1981). 13. H.
G. Muller, A. Tip, and M. J. vander Wiel,/. Phys. B 16,L679(1983). 14. T. F.
Gallagher, W. Sandner K. A. Safinya, and W. E. Cooke, Phys.Rev.A 23,
2065(1981). 15. M.Pimbert, /. Phys. (Paris) 33, 331 (1972). 16. T. F.
Gallagher, S.A. Edelstein, andR. M.Hill, Phys. Rev.A 14,2360(1976). 17. F.
Gounand, /. Phys. (Paris) 40, 457(1979). 18. P. R. Koch, H. Hieronymus, A.
F.J. van Raan, and W. Raith, Phys. Lett. 75A, 273 (1980). 19. G. F.
Hildebrandt, E.J. Beiting, C.Higgs, G.J. Hatton, K.A. Smith, F. B. Dunning,
and R. F. Stebbings, Phys. Rev. A 23, 2978 (1981). 20. W.P. Spencer, A. G.
Vaidyanathan, D. Kleppner, and T. W.Ducas, Phys.Rev.A 25, 380 (1982).

References 69 21. W. P. Spencer, A. G. Vaidyanathan, D. Kleppner, and T. W.

Ducas, Phys.Rev. A 26, 1490 (1982). 22. E. M. Purcell, Phys.Rev. 69, 681
(1946). 23. D. Kleppner, Phys.Rev. Lett. 47, 233 (1981). 24. A. G.
Vaidyanathan, W. P. Spencer, and D. Kleppner, Phys.Rev. Lett. 47, 1592
(1981). 25. J. M. Raimond, P. Goy, M. Gross, C. Fabre, and S. Haroche,
Phys.Rev. Lett. 49, 117 (1982). 26. R. G. Hulet, E. S. Hilfer and D.
Kleppner, Phys.Rev. Lett. 55, 2137(1985). 27. R. G. Hulet and D. Kleppner,
Phys. Rev. Lett. 51, 1430 (1983). 28. L. Holberg and J. L. Hall, Phys.Rev.
Lett. 53, 230 (1984). 29. W. M. Itano, L. L. Lewis, and D. J. Wineland,
Phys.Rev. A IS, 1233 (1982). 30. F. Gounand, P. R. Fournier, J. Cuvellier,
and J. Berlande, Phys. Lett. 59A, 23 (1976). 31. T. F. Gallagher and W. E.
Cooke, Phys. Rev. A 20, 670 (1979). 32. K. Bhatia, P. Grafstrom, C.
Levinson, H. Lundberg, L. Nilsson, and S. Svanberg, Z. Phys. A 303, 1
(1981). 33. M. Gross, P. Goy, C. Fabre, S. Haroche, and J. M. Raimond,
Phys.Rev. Lett. 43343 (1979). 34. F. Gounand, M. Hugon, P. R. Fournier, and
J. Berlande, /. Phys.B 12,547(1979). 35. A. G. C. Mitchell and M. W.
Zemansky, Resonance Radiation andExcited Atoms (Cambridge University
Press, New York, 1971). 36. R. C. Stoneman and T. F. Gallagher, Phys. Rev.
Lett. 55, 2567 (1985). 37. D. Kleppner and T. W. Ducas, Bull.Am.
Phys.Soc.21, 600 (1976). 38. R. M. Hill, and T. F. Gallagher, USPatent
4,024,396 (1977). 39. J. A. Gelbwachs, C. F. Klein, and J. E. Wessel, IEEEJ.
Quant Electronics QE-14, 77 (1978). 40. H. Figger, G. Leuchs, R.
Straubinger, and H. Walther, Opt. Comm.33,37 (1980). 41. T. W. Ducas, W.
P. Spencer, A. G. Vaidyanathan, W. H. Hamilton, and D. Kleppner, Appl.
Phys. Lett. 35, 382(1979). 42. P. Goy, L. Moi, M. Gross, J. M. Raimond, C.
Fabre, and S. Haroche, Phys.Rev.A 27, 2065 (1983).

6 Electric fields The effect of an electric field on Rydberg atoms, the Stark
effect, provides an interesting example of how the deviation from the
coulomb potential atsmall rin, for example an alkali atom, significantly alters
its behavior in a field from hydrogenic behavior. The difference is somewhat
surprising sincefieldeffects are fundamentally long range effects, and this
difference alone would make the Stark effect worthy of study. However, in
addition to its intrinsic interest, the Stark effect isof great practical
importance for the study of Rydberg atoms. Hydrogen We begin
byconsidering thebehavior of the H atom in a static field, ignoring the spin of
the electron. We start with the familiar zero field nlm angular momentum
eigenstates with the immediate goal of showing that important qualitative
features of the Stark effect are easily understood using familiar notions. If the
applied field E is in the z direction, the potential seen by the electron
isgivenby V=-- + Ez. (6.1) r Using the zero field n�m states we calculate the
matrix elements (ft�m|Ez|n'�'ra')ofthe Stark perturbation tothe zero field
Hamiltonian. Writing the matrix element in spherical coordinates and
choosing the z axis as the axisof quantization, {ntm\Ez\n't'm') =E{ntm\r cos
0\n't'm'). (6.2) From the properties ofthe spherical harmonic angular
functions we know that the Ez matrix elements are nonvanishing ifm! = m and
�'= ��1. The most obvious effect ofthe field is that itlifts the degeneracy
ofthe �ra statesof a particular n.If we neglect theelectric dipole coupling to
other n states, the Hamiltonian matrix has the same entry, �1/2/t2,for all the
diagonal elements and a set ofoff diagonal (�|�z|��l) matrix elements,
allofwhich are proportional to E. Ifwe subtract the common energy �l/2n2
wefindthat all entries of the matrix are now proportional to E. Therefore,
when the matrix isdiagonalized tofinditseigenvalues, they must 70

Hydrogen 71 all be proportional to E, i.e. the eigenvalues are given byXXE.

Stated another way, the H atom exhibits a linear Stark shift. Since the
eigenvalues are all proportional to E, theeigenvectors are independent of �,
and the Stark states are thus field independent linear combinations ofzero
field � statesofthe same n and m. The fact that the Stark states have linear
Stark shifts means that they have permanent electric dipole moments. Stark
states which have positive Stark shifts have the electron localized on the
upfield side of the atom andStark states which have negative Stark shifts have
the electron localized on the downfield sideof the atom. Since mis agood
quantum number, each set ofm states is independent ofthe others. We consider
first the circular \m\ = n � 1 states. They have no first order Stark shift since
there are no other states of the same m and n. There are, however, two m=n
� 2 states, which exhibit small linear Stark shifts. The Stark shifts are not
large since the radial matrix element is small,1 _ ^ w \/n2 � f2 (nt\r\nt+l)=
(6.3) The minus sign isfor radial wavefunctions Rne(r) ~ + / as r �> 0. On
the other hand, the m = 0 states also include the low � statesforwhich the
radial matrix element is large, and the extreme m = 0states have Stark shifts
of approximately �3n2E/2. The most straightforward wayto treat the Stark
effect is to use parabolic coordinates, for in parabolic coordinates the
problem remains separable even with the electric field.1"3 The parabolic
coordinates aredefined in terms of the familiar Cartesian and spherical
coordinates by (6.4) 0 = tan"1 y/x, and (6.5) Surfaces of constant � or rj are
paraboloids of revolution about the zaxis. From Eqs. (6.4) it is apparent that
� = 0 corresponds to the �z axis, and � = o� corresponds to r �� oo?
0# 0. Correspondingly, rj = 0 corresponds to the +z axis and?/-� oo
corresponds tor ^ oo? ^ ^ 0.Whenthefieldispointingin the zdirection,
corresponding to Eq. (6.1), the electron escapes to rj = oo. Equivalently, the
motion in the � direction is bound, but the motion in the rj direction is not. In
parabolic coordinates the Schroedinger equation for anelectron orbiting a
singly charged ion with an external field in the z direction, i.e. in the
potential given by Eq. (6.1), is written using Eqs. (6.5) as1"3 * = r + rj = rx
yz z = r(l + cos 0) z = r(l � cos 6) = tan"1 y/x, = Vf?7 cos 0 = V|?7 sin 0

72 Rydberg atoms r_v2 2 where 2 , . . . � . , r . > ( 6 . 6 ) and Wis the

energy. If we assume that solution can be written as the product,1"3 n�,V,
<l>) = �i(�)�2fo)eim*- (6.7) with m an integer or zero, and substitute Eq.
(6.7) into Eq. (6.6), we find two independent equations for ux(%) and u2(rj)
and ^ d M J ^ ^ ^) = 0. (6.8b) d^ \ d/y / \ 2 477 4 From Eqs. (6.8a) and (6.8b)
it is apparent that the sign of m is unimportant, and that the wavefunctions for
�m are degenerate, as might be expected for this cylindrically symmetric
problem. In Eqs. (6.8a) and (6.8b) the separation constants Zx and Z2 are
related by1'2 Z1 + Z2=\. (6.9) Zx and Z2 may be thought of as the positive
charges binding the electron in the � and rj coordinates. This point will
become more apparent shortly. Note that the two differences between Eqs.
(6.8a) and (6.8b) are that the field enters with different signs in the two
equations and that Zx and Z2 may have different values. The classic way of
determining the energies of hydrogenic levels in a field is to solve the
zerofieldproblem inparabolic coordinates and calculate the effect of the field
using perturbation theory. The zero field parabolic wavefunctions obtained
by solving Eqs. (6.8a) and (6.8b) have, in addition to the quantum numbers n
and |ra|, the parabolic quantum numbers nx and n2, which are nonnegative
integers.1 nx and n2 are the numbers of nodes in the ux and u2 wavefunctions
and are related to n and \m\ by n = nx + n2 + \m\ + 1. (6.10) They are in
addition related to the effective charges Zx and Z2 by (6.11a) and W| + ^
(6.11b)

Hydrogen 73 n1-n2=-5 n1-n2=1 n1-n2=5 Fig. 6.1 Charge distribution for H,

for "parabolic" eigenstates: n = 8, ra = 0, /% � n^ = �7 to 7. The dipole
moments that give rise to the first order Stark effect are conspicuous (from
ref. 4). The wavefunctions are given in terms of associated Laguerre
polynomials, and using the approximate form of the Laguerre polynomials for
large arguments, we can write an approximate, unnormalized wavefunction
as1 Taking the squared absolute value of Eq. (6.12) and using the definition
of the parabolic coordinates given in Eq. (6.4), we can write an expression
for the electron probability distribution in spherical coordinates,1 I</W2m|2
= ^ ~ 2 ( 1 + COS 602"i+H ( l - COS Qfn2+\m\e-2r/nt (6.13) Eq. (6.13)
shows explicitly that the nx � n2 � n wavefunctions are localized along the
+z axis while the n2 � nx~n wavefunctions are localized along the �z axis.
Similarly, nx � n2 ~ 0 wavefunctions are localized near the z = 0 plane. In
Fig. 6.1 we show the n = 8, m = 0, n1 � n2 = -7 to 7 wavefunctions to
illustrate the polarization along the field direction.4 Careful inspection also
reveals that the

74 Rydberg atoms are1 nodes lie on parabolas. Using the

zerofieldwavefunctions the firstorder energies _ -n2)n. (6.14) LTT 2 While
there is no explicit m dependence, there is an implicit m dependence due to
the fact that nt + n2 + m + 1 = n. For m =0, allowed values of nx � n2 are n
� 1, n � 3 . . ., �n + 1 while for m = 1 they are n � 2, n � 4 .. . �n+ 2.
The even and odd m levels are interleaved. Furthermore, for the quantum
numbers n and m there are n � \m\ Stark levels. Note that for the circular
states, \m\ =n � 1,nx = n2 = 0, and the first order Stark shift vanishes, as we
have already seen. By taking into account the matrix elements off diagonal in
n, the second and higher order contributions to the Stark effect can be
calculated. If the calculation is carried through second order, the energies are
given by1 Wnnin2m = ^\^-^{ni-n2)-^-nA [11n2- 3(^ - n2)2 - 9m2 + 19].
(6.15) In Z 16 The second order shift is always to lower energy, as expected
from oscillator strength sum rules.5 Equally important, the second order shift
breaks the m degeneracy. The energies can be calculated to higher order,6'7
but for many applications the first and second order shifts are adequate. This
point is made explicitly by Fig. 6.2, a plot of the H \m\ = 1 energy levels
from n = 8to n=14 in electricfieldsfrom 0 to 2 x 10~5(0to 105V/cm). Fig. 6.2
makes several important points about the Stark effect in H. First, the levels
exhibit apparently linear Stark shifts from zero field to the point at which
field ionization occurs at significant rates, shown by the broken lines in Fig.
6.2. Only when looking along the diagram from the side of the page is the
curvature of the levels evident. Second, the Stark levels of adjacent n cross;
there is no coupling, at least at this level of resolution. Finally, the red, or
down shifted, Stark states ionize near the classical ionization limit, but the
higher energy blue, or upshifted, states ionize only at higherfields. As shown
by Fig. 6.2, the Stark shifts are quite linear, except at the highest fields
shown, and the first order energies of Eq. (6.14) are adequate for many
purposes. On the other hand, even the second order energies of Eq. (6.15) are
not adequate for comparisons to precise measurements of the energy levels in
field. For example, using a fast H beam Koch8 observed the (10,8,0,1) ��
(25,21,2,1) and (10,0,9,0) -> (30,0,29,0) transitions in fields of 2.514 and
0.689 kV/cm respectively using the 10jum R24 and P24 CO2 laser lines.
Here the states are denoted (n,nun2,\ni\). From his results it is apparent that
the energy of the red shifted (30,0,29,0) state isgiven accurately
byperturbation theory results at about tenth order. In contrast, the blue shifted
(25,21,2,1) state is never given to the same accuracy, and, in fact, the
perturbation theory result does not converge but oscillates about the
experimental result with the minimum amplitude of the oscillations at about
tenth order.8

Hydrogen 75 n=8 0.5 1.0 1.5 2.0 E(105 a.u.) Fig. 6.2Stark structure
andfieldionization properties ofthe \m\ = 1 states ofthe H atom. The zero field
manifolds are characterized by the principal quantum number n.
Quasidiscrete states with lifetime r > 10~6 s (solid line), field broadened
states 5 x10"10 s < r < 5 x 10~6 s(bold line), andfieldionized states r < 5 x
10"10 s(broken line). Field broadened Stark states appear approximately
only for W> Wc. The saddle point limit Wc =-2\/E is shown byaheavy curve
(from ref. 3). The second point made by Fig. 6.2, although not with very high
resolution, is that the levels of n and n + 1cross. The extreme m � 0 red and
blue states have Stark shifts of approximately �3n2E/2y which combined
with the n - (n + 1) energy spacing of 1/rc3,yields a crossing field of
E=V3n5. (6.16) This field is associated with the Inglis-Teller limit, where
Stark broadening in a plasma causes levels of adjacent n to become
unresolvable.9 The fact that the red n + 1 and blue n levels actually cross, in
spite of the fact that they have the same ra, is a result of the fact that the use of
the parabolic coordinates diagonalizes the Runge-Lenz vector, which has a
different eigenvalue in the two crossing states. The fact that both the Runge-
Lenz vector and angular momentum are conserved in H in zero field was
exploited by Park10 to show in an elegant way that the transformation
coefficient between the parabolic nn1n2m states and the spherical

76 Rydberg atoms nim states wassimply aClebsch-Gordon coefficient. If

weexpand the Stark state \nn1n2m)as \nn1n2m) = \ | n(m)(n�m\nriin2m), then
the transformation coefficient is given in terms of Wigner 3J symbols as
(nnxn2rn\ntm) = (_i)(i 2 2 ' (6.18) m + nx � n2 m � nx + n2 An equivalent
form isgiven by Englefield.11 It is possible to find quite a variety of phases
for the transformation coefficients of Eq. (6.18).10"13 The phase depends on
the phase conventions established for the spherical and parabolic states. The
choice of phase in Eq. (6.18) is for spherical functions with an r�, as
opposed to (�r)e, dependence at the origin and the spherical harmonic
functions of Bethe and Salpeter. A few examples of the spherical harmonics
are given in Table 2.2. The parabolic functions are assumed to have an
(�n)'m'/2 behavior at the origin and an elm<p angular dependence. This
convention means, for example, that for all Stark states with the quantum
number m, the transformation coefficient (nnxn2rn\nmm)is positive. To the
extent that the Stark effect is linear, i.e. to the extent that the wavefunctions
are the zero field parabolic wavefunctions, the transformation of Eqs. (6.17)
and (6.18) allowsus to decompose a parabolic Stark state in a field into its
zero field components, or vice versa. Harmin14 derived an approximate
form of the transformation of Eq. (6.18) which is particularly useful. The
information contained in the quantum numbers nx and n2 always appears as
�(n1 � n2) and can also be represented by Zx and Z2, the separation
constants, a notion which easily passes into the regime in which n2 is not a
well defined quantum number. (In a field n1 is always a good quantum
number.) Explicitly, Harmin showed that Eq. (6.18) can also be written as14
(nnxn2rn\ntm) = I- (-1)�P�m (Z, - Z2), (6.19) where P�m is a normalized
associated Legendre polynomial of Bethe and Salpeter.1 It is defined for m <
0, and Jl1P2�m(x)dx = I.1 Eq. (6.19) is only defined over the region -1 <
Z1 � Z2<1, which is, however, a less restrictive requirement than that both
nx and n2 quantum numbers have integral values. We may also write Eq.
(6.19) in terms of the parabolic nx and n2 quantum numbers using Zj � Z2 =
(nx � n2)ln from Eq. (6.11).From Eq. (6.19) it is easyto seehow the
parabolic states are composed of the ntm states. From either the properties of
the 3Jsymbolorthe fact that PooOO = 1/V2 itis apparent that thens stateis
evenly

Hydrogen 77 -0.3 Fig. 6.3 Plots of the projection of the n = 100 Stark states
of m =0,1,and 2 on the n = 100 zero field d states, <100 n1n2m\l00 2m> for
m = 0, 1, and 2: m = 0 ( ), m = 1( ) m =2 (. . .).The values are obtained from
Eqs. (6.11) and (6.19). spread over the nn1n20 states. The �> 1 states are
moreinteresting. In Fig. 6.3 we plot Eq. (6.19) for �= 2 \m\ = 0,1, and 2.
From Fig. 6.3 it isclear that the m = 0 states at the edge of the Stark manifold,
i.e. those for which \n1 � n2\ ~ n, have much more d character than those in
the center. On the other hand, the \m\ � 2 states at the center of the Stark
manifold have substantial d character while the states at the edges do not. For
n = 10 and \m\ = 2 the largest value oi\Z1 � Z2 = \ni ~ ^21/n is 0.7, but at
higher n it approaches 1, at which point the \m\=2Stark states at the edge of a
Stark manifold have vanishingly small d character. Working out the parabolic
wavefunctions in terms of the Laguerre polynomials is useful in the analytic
calculation of the Stark effect using perturbation theory. However, it is not
useful in very strong electric fields. Here we outline a more general
procedure which is valid in strong fields and lends itself to numerical
computations. If we replace u^g) and u2(rj) in Eqs. (6.8a) and (6.8b)
by1"3'15 ^ (6.20a) (6.20b) and introduce a normalizing factor of V2JT to
conform to common usage, the wavefunction is given by which leads to the
separated equations written as ^ J | + *i [W _ y ^ = 0 (6.22a) and

78 Rydberg atoms :0, (6.22b) where v <a-2(-f+!ir+f)- <6 -23a> and (^ ^ f )

(6.23b) \ j Arj1 4 The%! and#2wavefunctions describethemotion
ofparticlesoftotalenergy W/4 in the potentials V(�)/4 or F(?/)/4. It is also
interesting to note that for the case E=0 the wave equations of Eqs. (6.22a)
and (6.22b) are similar to the radial equation for the coulomb potential in
spherical coordinates. Explicitly, making the substitutions > W/2 ^ m 2 - l
(6.24) in Eq. (2.10) for p(r) = rR(r) yields Eqs. (6.22a) and (6.22b). When E
= 0, thepotential V(rj) with Z2 substituted for Zx is the same asV(^). It
isapparent that for smallvalues of Z1 and Z2the potentials V(if) and V(rj) are
not as deep as for the higher values of Zt and Z2. Thus at the same energy W
the number of nodes which occur in the wavefunction increases with Zx or
Z2. Inspecting Eq. (6.24), we can see that when E ^ 0 that V(�) and V(rj)
are no longer identical for the same values of Zx and Z2. The motion in the
� direction is bounded for all energies since V(g) �> Et; as �^> ��.
Incontrast, themotion in the rj direction is unbounded, for any energy, since
V(rj) �� �Erj as rj�> oo. The different behaviors of V(g) and V(rj) as�
and 77^^00 lead to qualitatively different wavefunctions. The motion in the
�direction has awell defined integer quantum number, nl9 and the motion in
the rj direction is, in principle, a continuum, containing resonances. In
practice, at energies far below the saddle point in V{rj) the motion in the rj
direction also has a good quantum number, n2. In Fig. 6.4we show the
potentials V(t-)and V(rj) for �"=10~6, and \m\ = 1. The potentials are
shown for Zx and Z2 = 0.1, 0.5, and 0.9, corresponding to a blue shifted
Stark state, a middle state, and a red shifted state. Asshownby Fig. 6.4(a), the
�motion isalways bounded. InFig. 6.4(b)itis apparent that the saddlepoint in
V(rj) occurs at lower energy for the red state (Z2 = 0.9) than for the blue
state. Had we drawn the potential V(rj) for ra=0, it would be deeper at rj =0
and would lie generally lower than the \m\= 1 potential. For \m\ > 1 the
potential has a 1/rj2 centrifugal barrier at small rj and lies generally above
the m = 1 potentials of Fig. 6.4. To solve Eqs. (6.22a) and (6.22b) for E #
0we take advantage of the fact that the motion in �is bounded. This motion
always has a bound wavefunction and a

Hydrogen 79 0.003 0.002 0.001 0.000 -0.001 1000 2000 3000 4000 5000
6000 0.000 "��050 1000 2000 3000 4000 5000 6000 TI (a0) Fig. 6.4
Plots ofthe potentials V(g) and V(rj) in afieldE = 10~6 for \m\ =1. (a) V(g)
for Zx = 0( ), Zx = 0.1 (- - -), Zx = 0.5 ( ), and Zx = 0.9 (� � �) (b) V(rj)
for Z2 = 0 ( ),Z2 = 0.1( ),Z2 =0.5( ), andZ2 =0.9 (� ��)�Note thatthe
saddlepointin V(rj) occurs atlower energy for larger values of Z2. As rj ->
oo the Stark effect dominates both potentials, so that the � motion is always
bounded.
80 Rydberg atoms 1.5 1-0.5 IFig. 6.5 Eigenvalues Zx = 1 � Z2 asafunction
of energy for m = 0, E = 1.5 x 10~5au = 77 kV/cm, and �x = 0-20 (from ref.
15). well defined, integral quantum number, n1? which is simply the number
of nodes in the Xi(%) wavefunction. For any energy W and any value of nx
we can find the separation constant, or eigenvalue, Zx. A straightforward
way of finding the eigenvalues of Zx is to use the unnormalized WKB
approximation i.e.1'315 r cos I I Vr(jc)djt--1, (6.25) where T(�) =[W -
V(f;)]/2, and ^ is the inner turning point. The WKB wavefunction of Eq.
(6.25) satisfies the quantization condition (6.26) where nx is apositive
integer and �x and ^2 are the inner and outer classical turning points, where
W = V(�). An alternative approach is use a Numerov algorithm to find the
allowed wavefunctions, as discussed by Luc-Koenig and Bachelier.3 In
either case it is important to remember that for the same values of W, m, and
E there exist many solutions to Eq. (6.26) corresponding to different values
of the quantum number n^ and the eigenvalue Zx. These solutions are
orthogonal, which is why the H Stark energy levels of the same m cross, as
shown by Fig. 6.2. In Fig. 6.5 we show the eigenvalues Zx associated with
several values of the quantum number nx asafunction of the energy W. These
values of Zx are obtained by means of the WKB quantization condition of Eq.
(6.26). For a constant value

Hydrogen 81 of rii, the value of Zx decreases with increasing energy.

Keeping nx constant means the WKB integral of Eq. (6.26) must remain a
constant. The WKB integral is an integral of the momentum, or the square
root of twice the kinetic energy, over the classically allowed spatial region.
Increasing the energy increases both the spatial range covered by the integral
and the kinetic energy. To keep the same value of nx as W is increased, the
depth of the small � part of the potential is reduced, by reducing Zx. For
fixed W, m, and E there exists a series of values of n1 and Zx implying a
series of allowed values of Z2 = 1 �Z�. The known value of Z2 can now
be used to solve Eq. (6.22b). Since V(rj) �> �Erj/2 as rj �� o�, in
principle, the solution to Eq. (6.22b) is a continuum wave for all values of
the energy. Accordingly we normalize the %2{rj) wavefunction per unit
energy as rj �� o�.The asymptotic form of the Xiwavefunction is easily
obtained using the WKB approximation as3 M s i n [h(Er)+2W)3/2+a]' (6
-27) where a is a phase. In Eq. (6.27) we see the expected oscillatory
behavior and an amplitude decreasing as the square root of the momentum.
Normalization per unit energy is often obtained by requiring that the inward
or outwardflux/= \l2jt. An approach consistent with the one given in Chapter
2 is to require r rw+AW N2 U*dr i/>dW' = l. (6.28) J J W-AW In parabolic
coordinates the volume element dr = (�+?7)d�d?7d0/4. Using the general
form of the wavefunction given in Eq. (6.21) and carrying out the angular
integration, leads to , X*(t)X2(n){-+ \)^p-\ Xi(�)jfcfo)dW'. (6.29) 0 Jo W
S/ 4 JW-AW The major contribution to the integral of Eq. (6.29) comes from
rj�� o�,in which case the IIrj term may be ignored, leaving 9 f��(Xi*
(�ki(�)) f�� fW+AW ^yz // ^g xK7/)^ >feO7)dW = 1. (6.30) Jo � Jo J
W-AW Requiring that the bound %i(�) wavefunction be normalized
according to r~d�=l (6.31) Jo s yields (6.32) Thus, the amplitude of the
continuum Xiiv) wave varies insignificantly with energy as r �> 00.

82 Rydberg atoms (c) CD C LJJ Vfo) Fig. 6.6 Schematic wavefunctions

Xiijl) f�r several choices of energy W. (a) An energy below Vbthe peak of
the potential V(rj) butawayfrom aresonance ofthe inner wellofthe potential.
The wavefunction invanishingly small intheinner well of the potential, (b) An
energy below Vb and equal to one of the resonances of the inner well, (c) An
energy above Vb. We are not so much interested in the wavefunction at r =
�� as we are in the wavefunction in the inner potential well of V{rj).
Examining plots of the potential V(rj), it is apparent that there is, in general,
apeak Vbin the potential for rj > 0. If the energy W is above the peak of the
potential, W > Vb, the entire region is classically accessible. If the energy is
less than the peak of the potential barrier, Vb, there are two classically
allowed regions, which are connected by a classically forbidden region
through which tunnelling occurs. The ratio of the amplitude of the
wavefunction in the inner well to that in the outer well is the quantity of
interest. Consider for a moment the energy dependence of Xiiv) shown
qualitatively in Fig. 6.6. Outside the barrier it has the standing wave form of
Eq. (6.27) with virtually the same amplitude irrespective of energy. At an
energy below the peak of the barrier the small transmission through the
barrier allows some of the exterior wavefunction to penetrate into the inner
well. Due to the small transmission, in general the wavefunction in the inner
well of V(rj) is negligible, as shown by Fig. 6.6(a). However, at a resonance
of the inner well, when an integral number, n2, of nodes fit, the amplitude of
the wavefunction in the inner well becomes very large, as shown in Fig.
6.6(b). This situation is no different from a microwave cavity tenuously
coupled to a transmission line.

Field ionization 83 Far belowthe barrier in V(rj) the transmission through the
barrier isvanishingly small and the resonances are so sharp as to be bound
states for all practical purposes, i.e. n2 is, for all intents and purposes, a
good quantum number. As the energy isincreased the transmission through the
barrier increases, increasing the width of the resonances and decreasing the
amplitude of Xiiv) in the inner well. Finally, when the energy exceeds the top
of the barrier, the wavefunction has an amplitude which is spatially smoothly
varying, displaying minimal energy dependence asshown byFig. 6.6(c).
WhileFig. 6.6 showsthe variation of the amplitude of the wavefunction in the
inner well of V(rj) in a qualitative way, we need a quantitative measure. A
useful one takes advantage of the fact that for small values of �and rj the
regular solutions of #i(�)/Vf and XiivV^V behave as �'ml/2 and //'m'/2. A
useful approach is to define a small �, rj wavefunction as3 V = � ^ (^)|m|/2
em<t>, (6.33) V2JI where C�, the density of states, is a measure of the
probability of finding the electron near the origin, whichis ofpractical
interest for photoexcitation from the ground state and for field ionization of
nonhydrogenic atoms. C� is easily obtained from the normalized Xi(�) an
d Xiiv) wavefunctions. Figure 6.7 is a plot of Vc�, the square root of the
density of states for m = 0, nx = 7, and E = 1.5 x 10~5 au. Inthiscasethe
maximum inthepotential barrierof V(rj) is at Vh = -0.00372. Above Vb there
is, in essence, a continuum. Below Vb there is a series of sharp states which
each have the approximate quantum numbers n2 shown, corresponding to n2
nodes of the Xiiv) wavefunction in the inner potential well of V(rf). For
anym,eachvalueifthequantum number n1has a spectrum of VC� analogous to
the one shown in Fig. 6.7. Field ionization Field ionization, which is of great
practical importance in the study of Rydberg atoms,has been introduced
implicitlyin Figs. 6.2,6.6, and6.7.We nowconsiderit explicitly. The order of
magnitude of the field required for ionization can be estimated using the
method outlined in Chapter 1.The combined coulomb-Stark potential, V=-l/r
+ �z, (6.34) hasitssaddle point on the z axis atz = �\l\fE where thepotential
hasthevalue V = �2\fE. Thus if an atom is in a state of m = 0, so that there is
no additional centrifugal potential, and if the electron is bound by an energy
W, afieldgivenby W2 E =� (6.35)

84 Rydberg atoms 1x10 Fig. 6.7 Typical energy dependence ofthe square root
ofthe density ofstates VC� forthe states nx = 7, \m\ = 0in thefieldE = 1.5 x
10~5 au. Vh is the energy of the peak in the potential V(rj).Each resonance is
characterized by the parabolic quantum number n^ and the width Tn2
whichvariesrapidly withn2. ForW<Vb the spectrum exhibits quasi-discrete
resonances;for W> Vbthe continuum
exhibitsoscillationswhichdampoutwithincreasing n2. Theenergies andwidths
of the prominent resonancesare. Wx = -0.005066 Tx =3.7 x 10"16 au W2 =
-0.004097 T2 = 6.368 x 10"8 au W3 = -0.003601 T3 = 1.353 x 10"4 au W4
= -0.003100 T4 = 1.637 x 10"3 au (fromref. 3). is adequate for ionization to
occur classically. This field is usually termed the classical field for
ionization. If we ignore the Stark shift of a Rydberg state of principal
quantum number n and write the binding energy in terms of n we obtain the
familiar result for the classical field for ionization in terms of the principal
quantum number, E (6.36) 16n4 These results, Eqs. (6.35) and (6.36), are
only valid for m = 0states. In higher \m\ states there is a l/(x2 + y2)
centrifugal potential keeping the electron away from the zaxis, and the
centrifugal barrier raises the threshold field ofm^O states.16 Specifically, for
m # 0 states the fractional increase in the field required for ionization,
compared to an m = 0 state of the same energy is16 AE \m\Vw _ E V2 In
(6.37)

Fieldionization 85 Actually, the classical picture outlined above has two

serious defects. First, Eq. (6.36) ignores the Stark shifts, as does Eq.(6.37) if
we choose to use the \m\l2n form. Eq. (6.35) and the \m\VW/V2 form of Eq.
(6.37) do not suffer from this defect. Second, the classical approach ignores
the spatial distribution of the wavefunctions. As shown byFig. 6.2 for H
states of the same n and \m\, the higher energy, or blue, nx � n2 � n, Stark
states require higher fields to ionize than the lower energy, or red, n2 � n1 ~
n states. This point is made quite graphically by Bethe and Salpeter1 using
the data of Rausch von Traubenberg.17 A simple physical argument for why
blue states are harder to ionize than red states is that the electron is located
on the side of the atom away from the saddle point in the potential in the blue
states, whereas the electron is adjacent to the saddle point in the red states. In
parabolic coordinates the motion in the � direction is bounded. Thus for
ionization to occur the electron must escape to rj = oo.Classically, ionization
only occurs for energies above the peak Vh of the potential barrier in V(rj). If
we ignore the short range lit]2term in Eq. (6.23b), an approximation good for
lowm, and set W = Vh we find the required field for ionization to be W2 E =
�. (6.38) 4Z2 Eq. (6.38) differs from Eq. (6.35) by the factor of Z2, the
effective charge. Consider three low \m\states of the same energy. A blue
state of nx � n2 ~ n has Z2 ~ 1/n, a central state has nx = n2 � nil and Z2 ~
1/2, and ared state has n2 - nx ~ n and Z2 ~ 1. If the three states have the
same energy the red one will be most easily ionized. For the extreme red
Stark state of high n Eq. (6.38) reduces to Eq. (6.35) since Z2 ~ 1. For this
Stark state the Stark shift increases the binding energy, and for an m = 0 state
the energy is adequately given using the linear Stark effect as 1 3 � (6.39)
2nz Using this energy we find a threshold field 1 (6.40) The numerical factor
of 1/9 instead of 1/16 is due to the Stark shift of the level. For the blue states
it is not possible to estimate simply the threshold field. However, blue and
red states of the same n and m=0 often have threshold fields differing by a
factor of 2. We have thus far discussed field ionization as occurring only
when it is classically allowed, when E > W2/4Z2, i.e. when the energy is
above Vb, the peak of the barrier in V(rj), as shown in Fig. 6.6. At lower
energies tunnelling occurs, and accurate calculations of ionization rates have
been made.218"20 As an example, we show in Fig. 6.8 the ionization rates
calculated byBailey etal20 using

86 Rydberg atoms �i rrc 1 r1011 - O 101 108 107 106 8 E (kV/cm) Fig. 6.8
Calculated ionization rates of the (njix 9ri2,m) states asa function
offieldusing the values of ref. 20: the (15,0, 14,0) red state (O) and the
(15,14, 0, 0) blue state (�). the WKB method of Riceand Good18 for the H
n=15 extreme red and blue m = 0 Stark states over the range of ionization
rates from 106s"1 to 1011s"1. From Fig. 6.8 it is apparent that the blue states
ionize at higherfieldsthan red states of the same n. It isalsouseful to note that
the ionization ratesincrease rapidlywith field, implying that simply computing
when ionization is classically allowed is often quite adequate for practical
applications. Using a fast (8 keV) beam of H Koch and Mariani21 have made
precise measurements of the ionization rates of state selected \m\=0 and 1, n
= 30-40 blue Stark states of H. As can be seen in Fig. 6.9, the measured rates
are in quite good agreement with, but typically offset from, the analytic
results obtained by Damburg and Kolosov,22 using the approximation that the
field is weak enough that n2is very nearly agood quantum number. However,
as shown, the measured rates areinexcellent agreement withtheexact
numericalcalculations of Damburg and Kolosov,23 which are in good
agreement with the ionization rates calculated by Bailey et al.20 The
excellent agreement between theory and experiment suggests that a
nonrelativistic treatment is quite adequate. However, Bergeman has
calculated that it should be possible to see the effects of spin orbit coupling
in field ionization.24 The ionization measurements of Koch and Mariani20
are for blue states of nx ~ n. Rottke and Welge25 carried out measurements
on the relatively red \m\ =0 and 1 statesof n = 18 and bluestatesof n =
19verifying the rapid decrease in ionization rate with n1 shown theoretically
in Fig. 6.9. In their measurements

Nonhydrogenic atoms 87 0.21 n4E(au) 0.29 400 500 600 FIELD

STRENGTH (V/cm) Fig. 6.9 Dotted lines, experimental ionization rate
curves for the (n^^^ml) states 1: (40,39,0,0);2: (40,38,0,1);3: (40,38,1,0);4:
(40,37,1,1);5: (40,37,2,0);6:(40,36,2,1).The tick marks represent the range of
validity of the experiment. Solid line, theoretical curves calculated with Eq.
(6) of ref. 22.Squares, numerical theoretical resultsfrom ref. 23 (from ref.
21). they used pulsed laser excitation of a beam of H atoms in a field of 5714
V/cm followed by time resolved detection of the electrons from the ionizing
H atoms. Most of the discussion of field ionization thus far has been focused
on states of low \m\. For states of high \m\ we may no longer neglect the
centrifugal potential, and it isnot possible to derive a simple expression such
as Eq. (6.38). Nonetheless, since the centrifugal term raises the barrier in the
potential V{rj)it is clear that the ionizing field must be higher for high m
states. A graphic illustration of this point is provided by Fig. 6.10, a
simulation of the ionizing fields of all the \m\ > 3 states of n = 31.26
Measurements of the field ionization profiles of collisionally induced
mixtures of the degenerate high �,|ra| states of the same n are in substantial
agreement with the profile predicted in Fig. 6.10.26 Nonhydrogenic atoms
Atoms other than H have characteristics essentially similar to those of the H
atom in an electric field, but there are important differences due to the
presence of the finite sized ionic core. In zero field the presence of the core
simply depresses
Rydberg atoms -50 V70 I -110 '-130 -150 o e> o 3 �� 0 3CX) 600 900
1200 1500 ELECTRIC FIELD, VcrrT1 Fig. 6.10 Calculated SFI profile for
diabatic ionization of the H like |m| > 3 states. Top, extreme members of the n
= 31,\m\=3 Stark manifold. The crosses represent the points at which each
\m\ > 3 Stark state achieves an ionization rate of 109 s- 1 . Bottom,
calculated SFI profile for diabatic ionization of a mixture containing equal
numbers of atoms in each m :3 Stark level for n = 31at a slew rate of 109
V/cm s. (from ref. 26). the energies, particularly those of the lowest � states.
As long as the core is spherically symmetric, it does not alter the spherical
symmetry of the problem, and the effect is relatively minor. However, with a
finite sized ionic core the wavefunction is no longer separable in parabolic
coordinates. As a result the parabolic quantum number nu whichis agood
quantum number inH, isno longer good. The most important implication of nx
not being a good quantum number is that blue and red states are coupled by
their slight overlap at the core. In the region below the classical ionization
limit blue and red states of adjacent n do not cross as they do in H, but
exhibit avoided crossings as a result of their being coupled. Above the
classical ionization limit blue states, which would be perfectly stable in H,
are coupled to degenerate red states, which are unbound, and ionization
occurs rapidly compared to radiative decay. It is really an autoionization
process in which the blue state is coupled to the red continuum state at the
ionic core. In this chapter we shall treat the Stark effect in nonhydrogenic
atoms by methods which, while not the most powerful, provide physical
insight. We begin by calculating the energy levels of Na, as an example, in
the regime where the field ionization rates are negligible. Ignoring the spin of
the Rydberg electron, the Hamiltonian is given by V2 1 H= + - + Vd(r) + Ez,
(6.41) 2 r

440Nonhydrogenic atoms 89 540 - i i i ! 11 i i i | i r-n-rrii i | i rnj i i i i | i , i ,

j n . . | . . i i j i , i i | , 11 , , , i , i 0 1000 2000 3000 4000 5000 6000 FIELD
(V/cm) Fig. 6.11Calculated Na \m\ = 1energy levels (from ref. 27). where
Vd(r)isthedifference between the Na potential and the coulomb potential,
�1/r. Consequently, Vd(r) is only nonzero near r = 0. A straightforward and
effective method of calculating the energies and wavefunctions isdirect
diagonalization of the Hamiltonian matrix. If we use the Na ntm spherical
states as basis functions, the Hamiltonian has diagonal elements given bythe
known energiesof the zero field ntm states, i.e. (n�m\H\n(m) = l (6.42) 2(n -
dey and off diagonal matrix elements (n�m\Ez\n'� � Ira)which may be
calculated using the Numerov method outlined in Chapter 2. If several
nmanifolds of�states are included, theeigenvalues ofthe matrix
givetheenergy levelsofthe Na atomin the field, and the eigenvectors give the
Stark states in terms of the zerofieldn�m states. In Figs. 6.11 and 6.12we
show the energy levels for the Na states of n ~ 15 for \m\ = 1 and0,
respectively, obtained by direct diagonalization of the Hamiltonian matrix.27
In zero field the quantum defects 1.35 and 0.85, of the Na s and p states
displace them from the high �statesand they only exhibit large Stark shifts
when they intersect the manifold of Stark states. A second obvious difference
between Figs. 6.11and 6.12 andFig. 6.2is that the energy levels of different n
donot cross as they do in H. There are avoided crossings, which are clearly
visible for the \m\ = 1 states of Fig. 6.11 andoverwhelming for the ra =
0states of Fig. 6.12.The energy level diagram for the Na |ra| = 2 states would
not be observably different from the analogous diagram for H on thescale
ofFigs. 6.11 and 6.12. Atzero field

90 Rydberg atoms i I i i i i 1 i i i i I i i i iji i i i I i i i i I j MI 1 i ill I i I I I 1 I

I i I I I I I I 1 i n 440 11 i I p i i 11 11 i 111 11 i I i i i 11 i i i i I ii i i 0 1000
2000 3000 4000 5000 6000 FIELD (V/cm) Fig. 6.12 Calculated Nam = 0
energy levels (from ref. 27). the levels would be degenerate, and for E>
l/3n5 the blue and red levels of adjacent n would cross. A subtle point which
is not immediately apparent when inspecting Figs. 6.11 and 6.12 is that the
\m\ = 0 and 1 levels are not interleaved as they are in H. Note that in Fig.
6.12 the 16s state joins the manifold at a field of 2000 V/cm. At fields below
1500 V/cm the \m\ = 0 and \m\ = 1 levels of the lower half of the n = 15 Stark
manifold lie practically on top of one another. This point was demonstrated
explicitly by Fabre et al28 who used a diode laser to excite Na atoms from
the 3d state to states of the n = 29 Stark manifold with high resolution. In Fig.
6.13 we show the result of scanning the wavelength of the diode laser across
three high lying pairs of \m\ = 0 and 1 levels of the n = 29 Stark manifold in a
field of 20.5 V/cm. The n = 29and n = 30 Stark manifolds intersect at afieldof
83 V/cm, so the field of 20.5V/cm corresponds roughly to afieldof 550V/cm
inFigs.6.11 and 6.12, i.e. the 30p state has not yet joined the n = 29 Stark
manifold. The \m\ = 0 and 1 states are <200 MHz apart while the adjacent
pairs of Stark levels are ~2 GHz apart. In H the \m\ = 0 and 1levels would be
1GHz apart. The \m\ = 0 levels lie above the \m\ = 1 levels, i.e. they are
further displaced from the center of the manifold, or the zero field n = 29 high
�levels. Fabre etal.28 used aprojection operator technique to describe the
Stark shifts at fields below where low ( states of large quantum defects join
the manifold. A less formal explanation isasfollows. If, for example, the s
and p states are excluded, as inFig. 6.13below 800V/cm, effectively onlythe
nearly degenerate �> 2 states are coupled by the electric field. The only
differences among the \m\ = 0,1, and 2 manifolds occur in the angular parts of
the matrix element, i.e.1

Nonhydrogenie atoms 91 p=5 p=U p=3 Fig.6.13Part ofthe excitation

spectrum of theNan = 29 Starklevelsfrom the3d state in an
electrostaticfieldof 20.5V/cm corresponding to nx values of n � 3, n � 4,
and n � 5. The energy splitting between m =0(highest energy line inthe
doublets) and \m\ = 1 states is of the order of 180MHz. The arrows indicate
theoretical positions of energy levels obtained by a numerical
diagonalization of the Stark Hamiltonian (from ref. 28). � . (6.43) l) V }
Only in a few low � states is the m dependence significant, and as a result
the \m\ =0,1, and 2 energy levels are almost degenerate. However, due to the
low � states, the \m\ = 0 levels are displaced more than the \m\ = 1 levels,
which are displaced insignificantly more than the \m\ = 2 levels.
Reexamining Figs. 6.11 and 6.12, the avoided crossings of the blue and red
states of adjacent n are evident. When the avoided crossings are large they
can be measured optically as was done by Zimmerman etal.27 They excited
Li atoms in a beam from the 2s state to the 2p state and then to the 3s state by
two pulsed dye lasers at fixed wavelengths. The wavelength of a third laser
was scanned to drive atoms from the 3s state to Rydberg Stark states in the
presence of a static field. The transition to the Rydberg state was detected by
applying a high field pulse to the atoms subsequent to the laser pulses. The
ions were detected with a particle multiplier as the wavelength of the third
laser was scanned. In Fig. 6.14 we show their observation of the avoided
crossing of the (18,16,0,1) and (19,1,16,1) states. The experimental picture
of Fig. 6.14 is built up by scanning the third laser over the energy of the
avoided crossing for several fields near the avoided crossing field of 943
V/cm. As shown by Fig. 6.14, the observed energies of the levels match
those calculated by matrix diagonalization. It is also interesting to note that
the oscillator strength to the upper state vanishes at the anticrossing. At the
anticrossing the eigenstates have the form (1/V2) (tyA � tyB) where ^ A and
^ B are the eigenstates away from the crossing. As shown by Fig. 6.14, away
from the avoided crossing both states apparently have similar amounts of p
character, and at the crossing the upper state loses it all to the lower state.
The method used by Zimmerman et al.21 to measure the Li avoided crossing
shown in Fig. 6.14 requires that the resolution of the laser be finer than the
size of

92 Rydberg atoms -(2V/cm) Field Fig. 6.14 Level anticrossing in Li.

Intersection of the (rc^^Jml) states (18,16,0,1) and (19,1,16,1) at 321.5cm"1
and 943 V/cm. The calculated level structure issuperimposed on the data.
Note the vanishing oscillator strength to the upper level at the avoided
crossing (fromref. 27). the avoided crossing. An alternative approach used
byStoneman etal.29 bypasses thisrequirement. They observed the narrow
anticrossings between the K(n 4- 2)s states, which have a quantum defect of
2.18, and the low lying members of the n Stark manifold, which is composed
of �> 3 states. They excited atoms in abeam from the 4sstate to the 4p state
and then to the ns state with twopulsed dye lasers with linewidths of 1 cm"1.
Two tothree microseconds after the laserexcitation the atoms were
selectively field ionized, and only atoms which had undergone black body
radiation induced transitions to the higher lying (n + 2)p states were detected.
The (n + 2)p signal was recorded as the staticfieldwas slowly scanned
through the avoided crossing field over many shots of the laser. Away from
the avoided crossing only the (n + 2)s state was excited, and it could undergo
black body radiation induced transitions to the (n + 2)p state. The n Stark
state, composed of �> 3 states, cannot be excited from the 4p state or to the
(n + 2)p state. At the avoided crossing the eigenstates are 50-50mixtures of
the two states, and both are excited from the 4p state, with the result that the
same number of Rydberg atoms is always excited. However, each of these
two states has half the (n + 2)s-(n + 2)p black body radiation induced
transition rate, and the observed (n + 2)p signal decreases. An example of
their signals, the avoided crossing signals of the K 20s state with the lowest
lying \m\ = 0 and 1 states of the n =18 Stark manifold, is shown in Fig.
6.15.29 First we note the decrease in the 20p ion signal at each of the
avoided crossings, as described above. Second we note that the \m\ =0and 1
Stark states are nearly degenerate. If they wereinterleaved as in H, the two
avoided crossings would be separated by 50V/cm not 5 V/cm as they are in
Fig. 6.15. We have again encountered the noninterleaving of the energy

Nonhydrogenic atoms 93 < z: CD 400 402 404 406 408 410 ELECTRIC
FIELD (V/cm) Fig. 6.15 Anticrossing signal from the avoided crossing of the
K 20s state with thelowest energy n = 18 Stark manifold state. The left-hand
peak corresponds to the m = 0 anticrossing, and the right-hand peak tothe \m\
= 1 anticrossing (from ref. 29). levels of different ra, as shown in Fig. 6.13.
The spin orbit interaction, which we have ignored, isresponsible for the
avoided crossing between the 20s state and the \m\ = 1 Stark state.
Diagonalizing the Hamiltonian matrix, including the spin orbit interaction,
yields 450 and 47 MHz for the widths of the \m\ = 0 and 1 anticrossings of
Fig. 6.16, inreasonably good agreement with the measured values of 510
MHz and 80 MHz.29 In the heavier alkali atoms the spin orbit interaction
plays an even more dominant role.27 When the quantum defects are large and
\m\ is low enough that several states have large quantum defects, matrix
diagonalization using the ntm states as basis functions is probably the most
efficient approach. However, if the contributing quantum defects are small, a
different approach to Eq. (6.41), suggested by Komarov etal.30is useful.
They suggested usingthe hydrogenicparabolic states as a basis instead of the
zero field Mm states, in which case the Hamiltonian of Eq. (6.41) has
diagonal matrix elements (nn1n2m\H\nn1n2m) = - \ + Z -n{nx - n2)E + O(E2)
+ {nnxn2rn\V^r)\nnxn2m), (6.44) and nonzero off diagonal elements,
(nnxn2rn\H\n 'n[n2m) = <nn1n2'w|Vd(r)|n'n[n2m). (6.45)

94 Rydberg atoms CD 5 Field Fig. 6.16 Hydrogenic Stark energy levels (- - )

before considering theeffect of the core perturbation Vd(r). The alkali levels
after incorporating the diagonal and off diagonal matrix elements of Vd(r) ( ).
The diagonal elements shift the levels and the off diagonal elements lift the
degeneracy atthe level crossing.Note that the different shifts ofthe levels
move theavoided crossing from thefieldof the hydrogenic crossing. The
diagonal elements are the hydrogenic energies plus a small energy shift, and
the off diagonal elements are the couplings between levels. In Fig. 6.16 we
show two hydrogenic levels as dotted lines and the non hydrogenic levels as
solid lines. The shifts of the nonhydrogenic levels from their hydrogenic
counterparts are due to the diagonal matrix elements of Vd(r) and the avoided
crossings to the off diagonal matrix elements of Vd(r). It is straightforward to
evaluate the matrix elements of Vd(r) using the known transformation
between parabolic and spherical states given in Eqs. (6.18) and (6.19). If we
expand the parabolic states in terms of hydrogenic spherical states, the matrix
elements of Vd{r) may be written as30 (nn1n2m\Vd(r)\nrn[n^m) = Y
(nn^mlntm)(ntm\V (r)\A�m)(nr(m\nfnfm), e (6.46) where we have taken
advantage of the spherical symmetry of Vd(r) to require the same � on both
sides of the matrix element. We recall that the core interaction causes the
depression of the energy levels below their H analogues. Thus,
(n�m\Vd(r)\n�m) = ^ . (6.47) n This expression may be generalized by
realizing that Vd(r) is only nonzero for small r, where the dependence on n is
through the normalization factor n~312 Accordingly Eq. (6.47) can be
generalized to30 (n�m\V d(r)\n f �m) = - f ^ . (6.48)

Ionizationof nonhydrogenie atoms 95 Thus, (nnxn2rn\Vd\n'n[n2m) = >

(nn1n2m\ntm)� r^J^(n'tm\n'n{n2\m). (6.49) ^-� Vn n Since \(nnin2m\n(m)\2
is the nim fraction of the nn1n2m, state, it isevident that the depression of the
nonhydrogenic Stark levels below the hydrogenic value, givenbythe diagonal
matrix elements of Vd(r),is simplythe amount of each low � state it
possesses multiplied by its quantum defect. The nonhydrogenic energy must
be lower than the hydrogenic energy since <5�> 0for all �. At
thefieldwhere the diagonal energies of two levels, given by Eq. (6.43), are
equal, the off diagonal matrix element lifts the degeneracy. At the avoided
crossing the energy levels are split by kW=2\(nn1n2m\Vd(r)\n'n{nim)\, (6.50)
and the eigenstates are symmetric and antisymmetric combinations of the two
Stark states. Eq. (6.49) shows how the avoided crossings scale with n.
Averaged over �, (nnxn2m\n�m)2 must be equal to II{n - ra). Thus, for low
\m\ � n and n~ n', the matrix element of Eq. (6.49) is proportional to 1/n4,
and the avoided crossings exhibit a 1/n4 scaling. Ionization of nonhydrogenic
atoms A nonhydrogenic atom, Nafor example, differs from Hnot onlyinits
energylevel spectrum but also in howit isionized by afield.In particular, in a
nonhydrogenic atom there are two forms offieldionization. Thefirstisexactly
like H. A stateof approximate quantum number nx can itself ionize, keeping
the same value of nl9 in a field E if its energy W is enough for the electron to
surmount the potential barrier Vb in V(rj).Rates for thisform of ionization
increase rapidly inthevicinity of the threshold for classical ionization, just
asin H. In this form of ionization the blue states require higherfieldsfor
ionization than red states of the same energy. The second form ofionization is
similar to autoionization.31 In anonhydrogenic atom nx isnot agood quantum
number and bound states of high nx are coupled to Stark continua of low nx.
Thisform of ionization applies to those states other than the reddest Stark
states. The extreme red Stark states have nx =0 and ionize, as do the red H
states, at the classical ionization limit given by Eq. (6.35), modified by Eq.
(6.37) for m �^ 0 states. This point has been demonstrated explicitly by
Littman etal,32who measured the time resolved ionization of Na \m\ = 2
states subsequent to pulsed laser excitation. Their measured rates are in good
agreement with the rates obtained by extrapolation of the rates of Bailey
etal.20 shown in Fig. 6.8.

96 Rydberg atoms In H it is possible to have n1 > 1 states which are quite

stable against ionization even though energetically degenerate states of lower
nx have been converted to continua. In other atoms these blue nx > 1 states
are coupled by thefinitecore to the red continua and autoionize. A simple
physical picture isthat an electron ina blue orbit comes near the core once on
each orbit and has afiniteprobability of being scattered from the stable blue
orbit into the degenerate red continuum. A significant difference between
thisform of ionization and hydrogenic ionization is that its rate is not
exponentially dependent on the field. The existence of the second form
ofionization meansthat thereis alargeregime of field and energy in which
nonhydrogenic atoms have states which decay more rapidly by ionization
than by radiative decay, but are still spectroscopically narrow. InH
allstateshaveionization rateswhich areexponentially dependent on thefieldso
the analogous rangeis much smaller. Thefact that there exist sharp but
ionizing levels above the classical ionization limit but below the hydrogenic
ionization limit wasfirstobserved by Jacquinot et al. in Rb.33 It was most
clearly shown byLittman etal. in Li.31They excited abeam of Liatoms in
staticfieldsby twofixedfrequency lasers from the ground state 2s via the 2p
state to the 3s state. They then scanned the wavelength of a third laser to
drive transitions from the 3s state to Rydberg states havingnpcharacter. The
excitation to a Rydberg statewas detected byfieldionization of the atom and
subsequent detection of the ion with a particle multiplier. Spectra were
recorded at many staticfields,as shown by Fig. 6.17, which is composed of
spectra recorded with the third laser polarized perpendicular to the
staticfieldso as to excite only \m\ = 1 states. The spectra of Fig. 6.17 were
recorded in two different ways. The spectra of Fig. 6.17(a) were recorded
by applying a fieldionization pulse 3 jus after laser excitation and only
detecting the ions which result from ionization by the pulse, not spontaneous
ionization. The resulting spectra are then composed of states which are stable
for at least 3 jus against field ionization, or radiative decay, which is longer
than 3 jus for the states of Fig. 6.17. It isquite apparent that there are no
stable states above the classical ionization limit. Onthe other hand,
ifnofieldionization pulseis applied andtheionsfrom atoms
whichionizeinthefirst3jus after the laser excitation are detected, the result is
Fig. 6.17(c). The spectra begin at the classical limit and extend to much
higher static fields.What is most important to note is that levels appear over
afieldrange of typically afactor of 2with noapparent
changeinwidth,aphenomenon whichdoes not occur in H. Fig. 6.17(b) shows
the calculated H \m\= 1 energy levels. Inspecting Fig. 6.17(c) it is easy to see
that the levels disappear from the experimental spectrum when their decay
rates exceed the laser linewidth by a large margin. We may develop a more
quantitative understanding of nonhydrogenic ionization
bystartingwiththehydrogenic Stark states and addingthecore perturbation
yd(r), just as we calculated the sizes of the avoided level crossings. The
level crossings described earlier are the result of core couplings between the
bound

Ionizationof nonhydrogenie atoms 97 300 (o) 320 340 �tiii (a) (b) 340 ' "" 5
6 7 ELECTRIC FIELD (kV/cm) Fig. 6.17 Tunnelling and saddle point
ionization inLi. (a) Experimental map of the energy levels of Li \m\ = 1states
in a static field. The horizontal peaks arise from ions collected after laser
excitation. Energy is measured relative to the one-electron ionization limit.
Disappearance of a level with increasing field indicates that the ionization
rates exceed 3 x 105 s"1. The dotted line is the classical ionization limit
given byEqs. (6.35) and (6.36). One state has been emphasized by shading,
(b) Energy levels for H {n = 18-20, \m\ = 1) according to fourth order
perturbation theory. Levels from nearby terms are omitted for clarity.
Symbols used to denote the ionization rate are defined in the key. The tick
mark indicates the field where the ionization rate equals the spontaneous
radiative rate, (c) Experimental map as in (a) except that the collection
method is sensitive only to states whose ionization rate exceeds 3 x 105 s"1.
At high fields, the levels broaden into the continuum in agreement with
tunnelling theory for H (from ref. 32).

98 Rydberg atoms o w Fig. 6.18 The density of states C� for four different
values of the quantum number n1: n A> n B> n c> ni D' For e a c n value of nl
C�is composed of narrow resonances up the W = Vb, the energy of the
saddle point in V{rj). Above W = Vb C� x is continuous as shown. The
bound state A innx = nA can ionize into the nx = nc and nD continua in a
nonhydrogenic atom. There is often a peak in C� at the onset of the
continuous spectrum, as shown. nn1n2nm and n'n'^m states, couplings
between bound levels of different n1. The same coupling between a
hydrogenically bound ntm state and a continuum state of lower nx results in
ionization. Ignoring for the moment coupling between nominally bound states
of different �x, the ionization rate of the state nnxm to the en[mcontinua is
given by Fermi's Golden Rule, r = 2JIY \{nnxm\Vd(r)\en[m)\2. (6.51) The
summation extends over allvalues of n[ which are continua at the energy and
field in question. Since Vd(r) is only nonzero near r = 0 the matrix element of
Eq. (6.51) reflects the amplitude of the wavefunction of the continuum wave
at r ~ 0. Specifically, the squared matrix element is proportional to C^, the
density of states defined earlier and plotted in Fig. 6.18. From the plots of
Fig. 6.18 it is apparent that the ionization rate into a continuum substantially
above threshold isenergy independent. However, as shown in Fig. 6.18, there
is often a peak in the density of continuum states just at the threshold for
ionization, substantially increasing the ionization rate for a degenerate blue
state oflargernx. Thisphenomenon has been observed experimentally by
Littman etal?2 who observed a local increase in the ionization rate of the Na
(12,6,3,2) Stark statewhereit crosses the 14,0,11,2 state, at a field of 15.6
kV/cm, as shown by Fig. 6.19. In this field the energy of the

Ionization of nonhydrogenie atoms 99 rCsec'1) 15.4 15.6 15.8 16.0 E(kV/cm)

Fig. 6.19 Ionization rates forthe (12,6,3,2) levelof Nainthe region ofthe
crossingwith the rapidly ionizing (14,0,11,2) level. The levels are specified
as (n,nx,n2,|w|). The solid line is a theoretical curve. The dashed line is
calculated neglecting another avoided levelcrossing at 17.3kV/cm (from
ref.31). (14,0,11,2) state is just below the top of the barrier in V(rj)and its
ionizing rate is 1010 s"1. With this low an ionization rate the (14,0,11,2)
wavefunction still has a large amplitude at the core, and the autoionization
rate of blue statesintoit is at its highest. Finally, it is useful to obtain an
estimate of the typical ionization rates in the region between the classical
ionization limit and the hydrogenic limit using Eq. (6.51). Each term in the
sum of Eq. (6.51) can be written as Knnx/nlVd(r)|e/iim)|2 = \(nnxm\^m)
(n{m\Vd\e�m) (eim\en[m)\2 (6.52) The maximum value of n[ in Eq. (6.51)
isthe value of^ for which Z2 = W2/4E, or Zx = 1 � Z2 = 1 � W2/4E. As
shown by Harmin,14 a continuum form of Eq. (6.19), valid for small radii, is
(e�m\en[m) = (-!)%�& - Z2). (6.53)

100 Rydberg atoms We arenow in a position to evaluate Eq. (6.51) using the
explicit form ofEq. (6.52) and replacing the sum over nx by an integral over
Zx � Z2. Thefirstterm of the matrix element ofEq. (6.52) isgiven by Eqs.
(6.11) and(6.19),the second one by the continuum modification of Eq. (6.48),
andthe third by Eq. (6.53). In addition to these substitutions we convert the
summation of Eq. (6.51) to an integral over Zx � Z2. According to Eq.
(6.38), the classical ionization limit occurs at E = W2/4Z2. Consequently,
thecontributing range of Z2 isfrom W2/4E to 1. Written interms of Zx � Z2
the integral corresponding to the ionization rate of Eq. (6.51) is given by
[/On \ "1 cin2/V fl~W2/2E (� HJ^H ftnfr-Z&KZi-Zi), (6-
54)whered^isthemagnitudeofthedifferencebetweend � and the nearest
integer. For example, d� = 0.85 implies 6^ = 0.15. Thefirstterm inEq. (6.54)
reflects the amount of�character intheinitial nnxm state, the second term is
thecouplingto the �partofthe continuum, andthe third, the integral, reflects
how much of the continuum is available having I character. Thevalue ofthe
integral ranges from zero at energies just above the classical ionization limit
toone when W= � V2E. Approximating (nn1n2m\n�m)2 by IIn and the
integral ofEq. (6.54) by1/2 we find the approximate result T~n�^ (6.55) n
Observations to date indicate that Na |m\ =0,1,and 2 Stark states of n ~
20above the classical ionization limit have widths of �30-90 GHz, 1-3
GHz, and 100 MHz,34 in rough accord with Eq. (6.55). In estimating
ionization rates we have ignored all interactions between bound states ofhigh
nx. These interactions also lead tolocal variations inthe autoionization rates
which canbe quite striking andcanbe understood in a straightforward way.
For a moment let us disregard the coupling tothe continuum and focus on two
nominally bound Stark states which have an avoided crossing. As we have
already discussed, at the avoided crossing the eigenstates are linear
combinations of the two Stark states. If away from the avoided crossing one
of the two Stark levels is rapidly ionizing and the other slowly ionizing, at
the crossing the ionization rates of the two eigenstates will most likely be
about the same andhalf the rate ofthe rapidly ionizing state. A more
interesting case is one in which the two Stark states have equal
autoionization rates. Ifthe coupling isto the same continuum inboth cases, then
at the avoided crossing the coupling will vanish in oneof the eigenstates and
double in the other one. The phenomenon is basically the same asthe
vanishingof the oscillator strength to the upper level atthe avoided crossing
ofFig. 6.14. If the two Stark states donothave precisely the same ionization
rates away from the avoided crossing, one of the ionization rates cannot
vanish at the avoided

References 101 0.02 0.01b 0.006 3.7 3.8 3.9 4.0 4.1 E (kV/cm) Fig. 6.20
Ionization width vselectricfieldfor the Na (20,19,0,0) level near
itscrossingwith the (21,17,3,0) levelfrom experiment (data points) andfrom
WKB-quantum defect theory (solid line). The levels are specified as
(n9nl9n2,\m\) Because the lineshapes are quite asymmetric (except for
verynarrow lines), thewidthinthisfigureis taken tobetheFWHM of the
dominant feature corresponding to the (20,19,0,0) levelinthe photoionization
cross section. For the narrowest line, experimental widths are limited by the
0.7 GHzlaser linewidth. Error limits are asymmetric because of the peculiar
line shapes and because of uncertainties due tothe overlapping \m\ = 1
resonance (from ref. 37). crossing. This phenomenon wasfirstobserved by
Feneuille et al.35 and discussed further byLuc-Koenig and LeCompte.36
However, it is shown most clearly in the data of Liu et al.31 who measured
the width of the Na (20,19,0,0) state in the vicinity of its intersection with the
(21,17,3,0) state at 3.95 kV/cm. As shown by Fig. 6.20, the width of the state
decreases from 2 cm"1 away from the avoided crossing to 0.02 cm"1 at the
avoided crossing. Correspondingly, the width of the (21,17,3,0) state
approximately doubles at the avoided crossing. References 1. H.A.Bethe and
E. A.Salpeter, Quantum Mechanics ofOne and Two Electron Atoms
(Academic Press, New York, 1957). 2. L.D. Landau and E. M. Lifshitz,
Quantum Mechanics (Pergamon Press, 1977). 3. E.Luc-Koenig and A.
Bachelier, /. Phys. B 13, 1743 (1980).
102 Rydberg atoms 4. D. Kleppner, M. G. Littman, and M. L. Zimmerman, in
Rydberg States ofAtoms and Molecules, eds. R. F. Stebbings and F. B.
Dunning (Cambridge University Press, Cambridge,1983). 5. U. Fano and J.
W. Cooper, Rev. Mod. Phys.40, 441 (1965). 6. H. J. Silverstone, Phys.Rev.
A 18, 1853 (1978). 7. R. J. Damburg and V. V. Kolosov, inRydberg States
ofAtoms and Molecules, eds. R. F. Stebbings and F. B. Dunning (Cambridge
University Press, Cambridge, 1983). 8. P. M. Koch, Phys.Rev. Lett. 41, 99
(1978). 9. D. R. Inglis and E. Teller, Astrophys. J. 90, 439(1939). 10. D. A.
Park, Z. Phys. 159, 155 (1960). 11. M. J. Englefield, GroupTheory and the
Coulomb Problem (Wiley Interscience, New York, 1972). 12. D. R. Herrick,
Phys. Rev. A 12, 1949 (1975). 13. V. L. Jacobs and J. Davis, Phys.Rev. A
19,776(1979). 14. D. A. Harmin inAtomic Excitation and Recombination
inExternal Fields, eds. M. H. Nayfeh and C. W. Clark (Gordon and Breach,
New York, 1985). 15. D. A. Harmin, Phys.Rev. A 24,2491 (1981). 16. W. E.
Cooke and T. F. Gallagher, Phys.Rev. A 17, 1226 (1978). 17. H. Rausch v.
Traubenberg, Z. Phys. 54, 307 (1929). 18. M. H. Rice and R. H. Good, Jr., /.
Opt. Soc. Am. 52, 239 (1962). 19. J. O. Hirschfelder and L. A. Curtis,/.
Chem.Phys.55, 1395 (1971). 20. D. S. Bailey, J. R. Hiskes, and A. C.
Riviere, Nucl. Fusion 5, 41 (1965). 21. P. M. Koch and D. R. Mariani, Phys.
Rev. Lett. 46, 1275 (1981). 22. R. J. Damburg and V. V. Kolosov, /. Phys. B
12,2637 (1979). 23. R. J. Damburg and V. V. Kolosov, /. Phys. B 9, 3149
(1976). 24. T. Bergeman, Phys.Rev. Lett. 52, 1685 (1984). 25. H. Rottke and
K. H. Welge, Phys.Rev. A 33, 301 (1986). 26. F. H. Kellert, T. H. Jeys, G. B.
MacMillan, K. A. Smith, F. B. Dunning, and R. F. Stebbings, Phys.Rev. A 23,
1127 (1981). 27. M. L. Zimmerman, M. G. Littman, M. M. Kash, and D.
Kleppner, Phys.Rev. A 20, 2251 (1979). 28. C. Fabre, Y. Kaluzny, R.
Calabrese, L. Jun, P. Goy, and S. Haroche, /. Phys.B. 17, 3217 (1984). 29. R.
C. Stoneman, G. Janik, and T. F. Gallagher, Phys.Rev.A 34, 2952(1986). 30.
I. V. Komarov, T. P. Grozdanov, and R. K. Janev, /. Phys.B. 13, L573 (1980).
31. M. G. Littman, M. M. Kash, and D. Kleppner, Phys.Rev. Lett. 41, 103
(1978). 32. M. G. Littman, M. L. Zimmerman, and D. Kleppner, Phys.Rev.
Lett.37,486(1976). 33. P. Jacquinot, S. Liberman, and J. Pinard, in Etats
Atomiques et Moleculaires couples a un continuum. Atomes et molecules
hautement excites, eds. S. Fenuille and J. C. Lehman (CNRS, Paris, 1978).
34. J. Y. Liu, P. McNicholl, D. A. Harmin, T. Bergeman, and H. J. Metcalf,
inAtomic Excitation and Recombination in External Fields, eds. M. H.
Nayfeh and C. W. Clark (Gordon and Brench, New York, 1985). 35. S.
Feneuille, S. Liberman, E. Luc-Koenig, J. Pinard, and A. Taleb,/. Phys.B
15,1205 (1982). 36. E. Luc-Koenig and J. M. LeCompte, inAtomic
Excitation and Recombination in External Fields, eds. M. H. Nayfeh and C.
W. Clark (Gordon and Breach, New York, 1985). 37. J. Y. Liu, P. McNicholl,
D. A. Harmin, I. Ivri, T. Bergeman, and H. J. Metcalf, Phys. Rev. Lett. 55,
189 (1985).

7 Pulsed field ionization Because it can be efficient and selective, field

ionization of Rydberg atoms has become awidely used tool.1 Often
thefieldisapplied asa pulse,with rise times of nanoseconds to
microseconds,2"4 and torealize thepotential offieldionizationwe need to
understand what happens to the atoms as the pulsedfieldrises from zero to the
ionizing field. In the previous chapter we discussed the ionization rates of
Stark statesinstaticfields.Inthischapter we consider how atomsevolvefrom
zero field states to the high field Stark states during the pulse. Since the
evolution depends on the risetime of the pulse, it is impossible to describe
all possible outcomes. Instead, we describe a few practically important
limiting cases. Although we are not concerned here with the details of how to
produce the pulses, it is worth noting that several different types of pulse,
having the time dependences shown in Fig. 7.1, have been used. Fig. 7.1(a)
depicts apulse which rises rapidly to aplateau. Atoms in afast beam
experience this sort of pulse when passing into a region of high homogenous
field. Fig. 7.1(b) shows a rapidly rising pulse which decays rapidly after
reaching its peak. While not elegant, such pulses are easily produced. For
pulse shapes such as those of Figs. 7.1(a) and (b) the ability to discriminate
between different states comes mostly from adjustment of the amplitude of
the pulse. Fig. 7.1(c) depicts alinearly risingfieldramp inwhich the
discrimination between different states comes from the time in the rising field
at which they are ionized. The definition of the ionization
thresholdfielddepends on the experiment. If thefieldpulse reaches its peak for
200 ns, then, if the field produces an ionization rate of 3.5 X106 s"1, 50%
ionization results, and this isa possible definition of the threshold field. (a)
(b) t (c) t Fig. 7.1 Typicalfieldpulses: (a) apulsewhich risesin time to
aflattoppedpulse; (b) apulse which risesto a peak then falls quickly ; (c) a
linearlyrising field. 103
104 Rydberg atoms -200 -300 ^ -400 o -500 * -600 -700 E (kV/cm) Fig. 7.2
Energy levels of the H n = 15,m = 0Stark levels. Thebroadening of the levels
corresponds to anionization rate of 106 s"1. The extreme red andbluestate
ionization rates are taken from the calculations ofBailey etal. (ref. 5), and
those ofthe intermediate states are simply interpolated. Hydrogen Let us
consider first the H atom. The essential question is, how does a H atom pass
from near zerofieldto the highfieldinwhich ionization occurs? If the atom is
initially in a zero field parabolic state, as might be formed by passage
through a foil, it keeps the same nnxn2m quantum numbers as the field is
increased to the ionization field, which depends on n,nx,n2, and m of the state
in question. To a reasonable approximation the ionizing field is near the
classical field defined by Eq. (3.45). For the red, nx ~ 0 states E =* l/9n4.
However, in the blue nx � n 2 states the required fields are higher. Fig. 7.2
is an energy level diagram of the hydrogenic n = 15,m = 0 Stark states. Each
Stark state simply follows its energy level as thefieldisincreased from zero to
the high ionizingfield.As shown byFig. 7.2 the blue state requires a higher
ionization field than the red state. In Fig. 7.2 the ionization fields shown
correspond to an ionization rate of 106 s"1. Thefieldsfor the extreme blue
and red states are taken from the calculations of Bailey etal,5 and thefieldsfor
intermediate states are simply interpolated. We have also not shown any
other n states since the Stark states of different n, which cross each other,
have different values of nl9 and do not interact, as described in Chapter 6.
Usually the H Rydberg states are not prepared in the zerofieldparabolic
states, but in the spherical states by, for example, optical excitation. In this
case, when the field is applied we would expect it to project the single ntm
state onto the degenerate nn1n2m states, each of which would then follow its
own path to

Nonhydrogenic atoms 105 Field Fig. 7.3 Schematic energy level diagram of
Na m = 0 states in afield. The displacement of the zero field ns and np states
from the approximately hydrogenic �> 2 statesis evident. In the region l/3n5
< E < l/16n4 the levels of different n have avoided crossings, and above E =
l/16n4the states ionize bycouplingto degenerate red continua of low nx. In
thisregion the levels are shown by dotted lines. In a slowly rising pulse
atoms follow the adiabatic levelsshown, and eachzerofield state is correlated
to one state, andthusone ionizing field, at E = l/16n\ ionization asshown in
Fig. 7.2. The precise projections onto the nn1n2tn states are given by Eq.
(6.19) and Fig. 6.3.6 For example, the 15s state would be projected equally
onto allthe n = 15 Stark states of Fig. 7.2, whilethe 15p m = 0statewould be
preferentially projected onto states at the red and blue edges of the manifold.
The H atom is a special case. Since the states of the same n and m are all
degenerate inzero field, no matter how slowlywe apply afieldweproject the
nlm states onto the nn1n2m states, i.e. the transition is always diabatic. On
the other hand, as long as thefieldrisesslowly compared to the Aninterval, a
H atom in an nx state remains in the samenx state and ionization
alwaysoccurs at the same field, irrespective of the risetime of the pulse.
Nonhydrogenic atoms Pulsed field ionizationof an alkali atom differs from
the description just given for H because of thefinitesizedionic core, or
equivalently, the nonzero quantum defects. There are three importanteffects.
First, the zerofieldlevelscan onlybe spherical nlm levels, not parabolic
levels. Second,in the E > l/3n5 regime there are avoidedcrossings of states of
different n. Third, ionization can occur at lower fields than inH. Specifically,
inH blue states have higher ionizationfieldsthan red states, but in an alkali
atom this is not the case due to nx changing ionization.

106 Rydberg atoms The easiest way of thinking about field ionization of, for
example, Na isto start with a staticfieldenergy level diagram such asFig.
7.3,which shows schematically the Nam = 0states. In lowfields�is a good
quantum number. What is a low field obviously depends on �. Intermediate
fields are those in which �is not good, but E < l/3n5. Highfields are those in
excess of l/3n5. AtfieldsE > W2/4ionization is classically allowed and
occurs by coupling to the degenerate red continua. In this region of Fig. 7.3
the levels are shown as broken lines. Finally, at very highfields, not shown in
Fig. 7.3, the states themselves ionize by tunneling, as in H. How ionization
occurs depends upon how quickly thefieldrisesfrom zero to the high field
required for ionization. Consider a slowly rising pulse. In this case the
passage from the zero field ntm states to the intermediate field Stark states is
adiabatic. The zero field ndm state slowly evolves into a single Stark state.
When the field reaches l/3n5, the avoided crossings with Stark states of the
same m but adjacent n are encountered. If the field is rising slowly, the level
crossings are traversed adiabatically and the atom remains in the same
adiabatic energy level. At each avoided crossing the atom passes smoothly
from one Stark state into another, as shown by the differing slopes of the
adiabatic energy levels on either side of an avoided crossing of Fig. 7.3.
Finally, when the field reaches E = W2/4 ionization occurs by coupling to the
underlying red Stark continua composed of red nx � 0, states of higher n.
Fig. 7.3 illustrates the most important features of adiabatic ionization. First,
the energy ordering of the zero field states is preserved as they pass to the
high ionizing field. Second, the adiabatic states which connect to the
zerofieldstatesof one n are trapped in the energy range -V2{n - 1/2)2 < W < -
l/2(n + 1/2)2. A way of representing the energy in the field relative to the
zero field limit is to use the effective quantum number in
thefield,3rcs,defined by W = � l/2ns2. From Fig. 3 it is apparent that the Na
(n + l)s and nd states pass to states of ns � n � 1/2 while the (n + 1) p states
pass to states of ns � n + 1/2, which are much more easily ionized. There is
only a small difference in the ionization fields of the nd and (n + 1) s states in
spite of their large zero field separation. In contrast, there is a large
separation between thefieldsrequired to ionize the nd and (n + l)p states in
spite of their zero field separation's being less than half the (n + l)s-nd
separation. If we assume that the energies in the high field are evenly spaced
and group the (n+ l)s and ni states together, we may estimate the ns values of
the series of m = 0 states as follows,3 (n + l)s->ws = n - 1/2 nd->ns = n-
1/2+ \ln ni^ns=ll2 + 2ln (7.1) nt^ns = n- 1/2 + (� - l)/n (n + l)p -> ns = n +
1/2 - 1/n.

Nonhydrogenic atoms 107 30 i25 20 LU DC � 15 o o ~ 10 0 I I 4.25 4.50

4.75 E (kV/cm) Fig. 7.4 Na 18s ionization threshold. The field ionization
current is plotted vs the peak ionizing field (from ref. 3). Using these values
of ns we can immediately compute the field required to ionize any m = 0
state, W2 1 E (7.2) 16n In adiabatic ionization what is important is the
ordering of the zero field energy levels, not their precise zero field energies.
In the Na atom the ionization of optically accessible states of n < 18 by 1 jus
risetime pulses is often adiabatic, and is described by Eqs. (7.1) and (7.2). In
Fig. 7.4 we show the ionization signal from the Na 18s state exposed to
a0.5JUS risetime pulse which is within 5% of its peak field for 200ns. As
expected, there is a single threshold, at 4.38 kV/cm, close to the 4.33 kV/cm
predicted by Eqs. (7.1) and (7.2). The width of the threshold, from 10 to
90% ionization, is �3%, asmight be expected for apulse of this shape. There
is asmall, 5%, signal below the threshold due to black body radiation driving
the atoms to higher states between the laser excitation and the field ionization
pulse.7 When field ionization threshold curves such as the one shown in Fig.
7.4 are measured for many states, they can be plotted together to exhibit the n
dependence of the ionization threshold field. In Fig. 7.5 we show a plot of
the threshold fields (50% ionization) for the Na \m\ = 0, 1, and 2 states
obtained with a 0.5 JUS risetime field pulse similar to the one shown in Fig.
7.1(b).8 In Fig. 7.5 it is apparent that, while the threshold fields of the m = 0
states are described by Eq. (7.2), \m\ = 1 and 2 states ionize at slightly higher
fields, which can be easily understood. Due to the centrifugal barrier, m ^ 0
states are excluded from the z

108 Rydberg atoms \�i 100 10 1 7 ' ' ^' ' ' \ A 1 1 1 1 1 D A \ 1 I 1 1 m| = 0 !
m| = 1 m| = 2 [ \ " 1 \ i i i i i i 6 8 10 12 14 16 18 20 ns Fig.7.5Plot of
observed ionizingfieldsfor Na |m| =0,1, and2 statesfor n = 8-20plottedvs ns,
the effective quantum number in the strong field. The line indicates the
classical ionization threshold (from ref. 8). axis, the location of the saddle
point in the potential. Thus the fields required to ionize m^ 0 states are higher
than the value given in Eq. (7.2). An approximate expression for the increase
in field required for ionization of a state with m ^ 0 above the field required
to ionize an m = 0 state of the same ns is9 AE/E = \m\l2n. (7.3) For n = 15,
\m\ = 1 and 2 states are predicted to ionize atfields3% and 5% higher than
their m = 0 counterparts, in reasonable agreement with the data shown in Fig.
7.5. It is useful to compare the difference between the thresholds of adjacent
\m states of the same � to the difference between adjacent t states of the
same m. From the assignments of Eq. (7.1) the fractional change in ionizing
field for an m = 0 state when � is increased by one, corresponding to Ans =
\lns, is AE/E = Alnl (7.4) At n = 20 changing � by one produces a change in
ionizing field of 1%, while changing \m\ from 0 to 1 produces a change of
2.5%. Comparing these two

Nonhydrogenie atoms 109 fractional changes to the width of the threshold

field shown in Fig. 7.4, it is apparent that it should be straightforward to
resolve adjacent \m\ states,but more difficult to resolve adjacent �states of
the same \m\. What conditions must be fulfilled for the ionization process to
occur in the adiabatic fashion we have just described? First, the transition
from the zero field n�m states to the intermediate field Stark states must be
adiabatic. Second, the traversal of the avoided crossings in the strong field
regime, E > l/3n5 must be adiabatic aswell. Finally, ionization only occurs at
E > W2/4 if the ionization rate exceeds the inverse of the time the pulse
spends with E > W2/4. Let us consider the first issue, the passage from zero
field to the intermediate field. The zero field spacing between � levels is
((5�- d�+1)/n3. When thefield reaches a value such that the separation
between Stark states, 3nE, is equal to the zerofieldspacing of the �states, the
�states are no longer good eigenstates. If the field reaches this value in
atime which islong compared to the inverse of the zero field splitting the
passage is adiabatic. On the other hand if the fieldreaches this valuein atime
short compared totheinverse of thezerofieldsplitting,thepassage is diabatic,
and the zero field n�m state is projected onto several Stark states. Assuming
the field rises linearly in time, we can write a useful criterion for
decidingifthe passagefrom lowtointermediatefieldis adiabaticor not in terms
of theslewrate S = dE/dt. Wedefine acritical slewrate S� for thepassagefrom
azero field �state by c (d<? ~ de+1 )2 In practice Eq. (7.5) means that n <
20states with quantum defect differences of 10~3 satisfy the adiabaticity
requirements for pulses with risetimes of 1 jus. On the other hand the high
�states have quantum defect differences which are much smaller than 10~3
and they do not satisfy the adiabatic criterion of Eq. (7.5) for the same
risetime. As a result, when thefieldis turned on they are projected
diabatically onto the intermediatefieldstates. From Eq. (7.5) it isclear that if
the risetime is kept constant, the � at which diabatic passage from zero field
occurs becomes lower as n is increased. Since it is impossible to excite
optically high � states, the statement that they pass diabatically from
lowfieldto the intermediate regime has not been tested, but it has been
experimentally established that the optically accessible low �states of n ~
20 Na atoms doin fact pass adiabatically to the intermediate field regime for
pulses with 1 jus risetimes. We nowconsider howthe avoided crossings in the
highfieldregime, E > l/3n5, are traversed. Consider an isolated avoided
crossing of magnitude a)0 between two levels 1, and 2 with different Stark
shifts dW^dE and dW2/dE, as shown in Fig. 7.6. If the atom is initially in
state 1 and the crossing is traversed slowly compared to l/a>0, the inverse of
the magnitude of the crossing, the traversal is adiabatic, as shown by the
solid arrow of Fig. 7.6. If the crossing is traversed rapidly compared to
1/CDQ the passage isdiabatic, asshown bythe broken arrow of
110 Rydberg atoms Fig. 7.6 Two Stark levels with an avoided crossing (o0.
If the field is slewed through the avoided crossing in a time long compared to
l/co0 the passage is adiabatic (solid arrow), while if it is slewed rapidly
through the crossing the passage is diabatic (broken arrow). Fig. 7.6. The
criterion for the critical slew rate for the crossings, Sx, may be written as
dWj _ dW2\ ~dE~~dEj (7.6) If S � Sx the avoided crossing istraversed
diabatically and if S � S x it is traversed adiabatically. Between the two
extremes of purely adiabatic and purely diabatic traversals, the probability
of making the diabatic transition is given by the Landau-Zener transition
probability, as has been demonstrated by Rubbmark et al.10 To apply Eq.
(7.6) we approximate the size of an avoided crossing by assuming that
(nnin2m\ntm) ~ \l\rn and only use the contribution from the lowest
contributing �state, � = m. In this approximation we obtain w0 - d'm Vn\
(7.7) where drm is the absolute value of the quantum defect d� (modulo 1).
Eq. (7.7) makes it apparent that the avoided crossings are traversed more
diabatically asn or \m\increases. ForNado = 0.35 d[ � 0.15(52 = 0.015. If
weuse these values for OJ0 in Eq. (7.6) with the largest possible value of
(dWt/dE - dW2/dE) = 3n2, we conclude that the level crossings should be
transversed adiabatically for pulses of risetime ~1jus even for n ~ 100 for
\m\ = 0,1, and 2 states. Above E = W2/4 the core coupling is between a
discrete state and a red Stark continuum, and the ionization rate F tothe
available continua is givenby Eq. (6.54).Ionization occurs at a field
dependent rate, and the time required for nearly complete, 90%, ionization
is2/F. One can estimate the ionization rates just above E = W2/4 to be F ~
d'm(2/n4) � (I/ft). The ionization rate is estimated assuming that one nx
channel

Nonhydrogenic atoms 111 is available, so that the integral of Eq. (6.54) is

~l/n. Using these rates one expectsto observe ionization in ~1 ns for \m\ =0
and 1 states and 30 ns for \m\= 2 states at n = 20. For most practical purposes
these times are fast enough that ionization appears to occur as soon as E >
W2/4. Contrary to the estimate above, for Na nd states of m =2 and n > 18
non adiabatic ionization is observed. Specifically, when Na nd states of m =
2 are exposed to 0.5 jus risetime pulses similar to Fig. 7.1(b), they exhibit
multiple ionization thresholds.3 There are, in principle, two possible
explanations. The first is that above the classical limit the atom passes
through regions in which the ionization rate varies enormously. If this were
the case we would expect to see occasional decreases intheionization
probability as afunction ofpulse amplitude, not a monotonic increase with
ionizing field. A more likely explanation for the multiple m =2 ionization
thresholds is that some of the avoided crossingsin the E > l/3n5 region
aretraversed partially diabatically, resulting in the atoms following several
paths to the classical ionization limit, resulting in several ionizing fields.
Support for this explanation comes from thework ofVialleand Duong11 who
used a stepped pulse to study fieldionization in Na, and of Jeys etal.4 who
traced the evolution from adiabatic to diabatic ionization in Na. Jeys et al.4
observed the transition from adiabatic to diabatic passage with increasing n.
They studied the ionization of the Na nd states of n > 30 with a linearly
risingpulse asshown inFig. 7.1(c), using the time atwhich electrons were
detected to determine thefieldin which the atoms were ionized. Their results
for nd states 30 < n < 36are shown in Fig. 7.7. Let usfocus on the 32d state.
Mostof the atoms ionize when the field reaches ~ 400 V/cm. A small number
of atoms ionizesbetween 450 and 600 V/cm, and a significant number of
atoms ionize at 600 V/cm. Ionization at �400V/cmcorresponds tofollowing
the adiabaticpath shown bythe broken line of Fig. 7.7(b). The \m\ =0and 1
states excited certainly follow this path. Ionization at 600 V/cm is due to \m\
= 2 atoms and corresponds to following the completely diabatic path shown
by the bold line in Fig. 7.7(b), the diabatic hydrogen-like path of the reddest
\m\ = 2 Stark state. Ionization at fields between 450and 600 V/cm is
presumably due to \m\ =2 atomswhich follow paths between the adiabatic
and diabatic paths of Fig. 7.7(b) and ionize at the classical limit. It is also
clear that the high field \m\ =2 feature grows in size as n is increased,
suggesting that, as n is increased, progressively larger fractions of the atoms
make the completely diabatic passage as shown by the bold lines of Fig.
7.7(b). The observation that diabatic passage is more likely with increasing n
is consistent with the onset of multiple \m\ = 2 thresholds at n = 18. It is also
important to notice that no ionization occurs atfieldsabove the \m\ = 2 feature.
Thus, no H-like states other than the extreme red ones are being populated.
The ionization of the Nandstates atn = 36 corresponds exactly tothe
ionization of the red Hn = 36,n1 = 0, \m\ =2 state,
whichionizeswhenclassically allowed at a field E � l/9n4. The form of
ionization exhibited by the \m\ = 2states is often termed diabatic ionization.
However, it is really only when E > l/3n5 that it is diabatic.
112 Rydberg atoms � 0 100 200 300 400 500 (b) ELECTRIC FIELD
STRENGTH , V/cm Fig. 7.7 (a) Field ionization data for Na nd states of n =
30,32,34,and 36. (b) Lightlines: extreme members of \m\ = 0 Stark manifolds
(fourth order perturbation theory); dotted lines: adiabatic paths to ionization
for n = 30, 32, 34, and 36; dark lines: diabatic paths to ionization for lowest
members of \m\ = 2 manifolds for n = 30, 32, 34, and 36. The lines indicating
the classical ionizationfieldsare calculated on the basis of Ref. 5 (from ref.
4).

Nonhydrogenic atoms 113 The passage from the zero field nd state to a single
Stark state is evidently adiabatic. A hydrogen nd |ra|=2 state would pass
diabatically from zero field to many nnin22 Stark states and exhibit multiple
ionization fields. Exposure of blue Stark statesto rapidly risingfields also
resultsindiabatic, or H like, ionization, at fields far above the classical field
for ionization. This point has been demonstrated byNeijzen and
Donszelmann12 usinghighlying,n = 66,states of In and by Rolfes et alP using
Na atoms of n = 34 and \m\ = 2. Apparently the Na \m\ = 2 states of n ~ 35
traverse the avoided crossings diabatically and do not ionize rapidly if they
are above the classical ionization limit, in clear contradiction to our earlier
estimates that all the avoided crossings for E > l/3n5 would be traversed
adiabatically, and ionization would occur at the classical limit E = W2/4.
What is wrong with the estimates? In our estimates of Ao>0 and T we
assumed that the d state was evenly spread over the \m\ = 2 Stark states, that
is (nn1n22\n22) ~ 1/Vn, irrespective of nt and n2. In fact, for the extreme
Stark states \nx + n2\ = n - 3, \(n(n- 3)02|n22)| = \{n 0n - 3 2| n22) = 6V5/n3/2
a factor ~n smaller than our estimate. The d states are concentrated in the
central Stark states where nx ~ n2 ~ nil, asshown byFig. 6.3.As aresult, the
avoided crossings between the extreme states, which are traversed most
rapidly, are a factor of n2 smaller than our estimate based on the average
value of the transformation coefficients. Therefore the risetime must be a
factor of n2 slower than our estimatesfor thepassage tobe adiabatic. At
theslew rates actuallyusedit is not surprising that the crossings are traversed
diabatically. Similarly, at the classical limit, the coupling of an extreme blue
Stark state to the underlying red continuum is reduced by a factor of n2. For
this reason the blue state does not ionize when it crossesthe classical
ionization limit, rather, onlywhen it reachesits own hydrogenic ionization
limit, as shown by Fig. 7.2. In the discussion of adiabatic ionization we have
implicitly assumed that this ionization occurs if the field exceeds the
classical ionization limit. As shown in Chapter 6, there are substantial
variations inthe ionization rates due to variations inthe density of statesof the
red continua and avoided crossings withmore rapidly ionizing levels.
Although the variations in the ionization rate above the classical limit are not
the cause of the multiple ionization thresholds observed in Na, they can lead
to nonmonotonic ionization curves,which have been observed in several
cases.814 A particularly clear example is shown in Fig. 7.8.14 Van de Water
et al. excited fast, 11 keV, He atoms to n = 19 triplet states in fields of 2.1-2.3
kV/cm, just below the classical ionization limit. The atoms then passed
adiabatically through a buffer field and into a field ionization region 8.24 cm
long. Subsequent tothe field ionization region was anionizingregion inwhich
allRydberg atomsnot ionized inthe ionization region wereionized. They
observed the number of atoms surviving the ionizing field as a function of its
magnitude near the classical ionization limit. As shown by Fig. 7.8, the
ionization curves are anything but monotonic, indicating the variation of the
ionization rate with fields in the continuum of the red Stark states. The atoms
are exposed to the ionizing field for

114 Rydberg atoms 2.4 2.5 Field (kV/cm ) Fig. 7.8 Signal of highly excited
triplet He atoms of the m = 0 manifold which survived exposure to the
ionizing electricfieldE as a function of its strength. The curves show the
ionization behavior of two different adiabatic states of ns = 19(from ref. 14).
200 ns, roughly the same time asfor the data shown in Fig7.4, yet show
markedly more structure. The clarity of the structure in the threshold of Fig.
7.8isprobably dueprimarily to the difference inthepulse shapes.The data
ofFig.7.8 were taken with the pulse shape of Fig. 7.1(a), while the data of
Fig. 7.4 were taken with the pulse shape of Fig.7.1(b). Multiplethresholds
occurbecause ofpartially adiabatic traversals of avoided crossings of bound
levels or incomplete ionization upon reaching the classical ionization limit.
These two causes need not occur separately, but can occur together, and
observations consistent with their occurring together have been made by
McMillian et al.15 It is useful topresent visuallythe difference between
adiabaticand diabatic field ionization. In Fig. 7.9 we show schematically
how adiabatic and diabatic ionizations occur for three n = 15 states. Diabatic
ionization, shown by the solid bold lines, is exactly like hydrogen. Onlythe
red state ionizes at the classical ionization limit; the fields for other states are
higher. In adiabatic ionization, shown by the bold broken lines, the n =
15levels are trapped between the n = 14 and n = 16 levels and ionize at the
classical ionization limit. In reality the true adiabatic levels, are not field
independent, as they are shown in Fig. 7.9, but exhibit the avoided crossings
shown in Fig. 7.3. However this simplification in the drawing

Spin orbit effects 115 300 5 10 15 20 E (kV/cm) Fig. 7.9 Adiabatic and
diabaticpathsto ionization for n = 15 statesinthe center and on the edges of the
Stark manifold. The diabatic paths are shown by solid bold lines and the
adiabaticpathsbybroken bold lines. Inboth casesionization occurs
atthelargeblackdots. The diabatic paths are identical to hydrogenic behavior.
The adiabatic n = 15 paths are trapped between the adiabaticn = 14 andn =
16 levels.Adiabaticionization alwaysoccurs at lower fields than diabatic
ionization. does not affect the reasoning. From Fig. 7.9 it is clear that
adiabatic ionization occurs at lower fields than diabatic ionization in all
cases. Spin orbit effects We have until now ignored the spin of the electron. It
has two effects; it splits the zero field� > 0 states, and it alters the avoided
crossings in highfields.In H, Li, and Na the second, high field, effect is
negligible so we shall for the moment ignore it. In H the fine structure lifts
the zero field � degeneracy and has, in principle, the effect of allowing
adiabatic passage from zerofieldspherical states to Stark
statesinthefield.However, since thefinestructure splittings arenot large
compared to the radiative decay rates, it does not appear that this possibility
is of much practical importance. In light alkali atoms, Li and Na, the fine
structure splitting of a low ( state is typically much larger than the radiative
decay rate but smaller than the interval between adjacent �states. In
zerofieldthe eigenstates are the spin orbit coupled isjrrij states in which
�and sare coupled. However, invery smallfields�and sare decoupled, and
the spin may be ignored. From this point on all our previous analysis of
spinless atoms applies. How the passage from the coupled to the uncoupled
statesoccursdepends onhowrapidly thefieldis applied. Itistypically a
simplevariant of the question of how the \m\ states evolveinto Stark states.
When

116 Rydberg atoms INTERMEDIATE FIELD Fig. 7.10 Adiabatic correlation

diagram for the Na nd states obtained from the known d state fine structure
splitting, the intermediate field energy ordering, and applying the nocrossing
rule for states of the same ra; (from ref. 3). the field reaches the point that the
Stark splitting of the \m\ levels is equal to the zero field fine structure
interval the states are uncoupled mms states. This occurs when a2E2 = WFS,
(7.8) where WFS is the fine structure interval and a2 is the tensor
polarizability. If this field isreached rapidly compared to the inverse of
thefine structure interval, the passage is diabatic, and the �syra;states are
simply projected onto the uncoupled �msms states, with the projections
given by Wigner 3J symbols or ClebschGordon coefficients. If this field is
reached slowly the passage is adiabatic, each (.sjnij state passing into one
�msms state. In an adiabatic passage the ordering of the states is critical as
in all adiabatic processes. Consider Na, for example. The zero field nd states
are inverted, and in thefieldthe uncoupled \m\ = 0, 1, and 2 states are ordered
as shown in Fig. 7.10. The m ordering is derived from the fact that the d-f
dipole matrix elements have magnitudes which decrease as \m\ is increased
from 0 to 2. In Fig. 7.10 wehave also shown the adiabatic correlation in
which rrij = m + ms isconserved.3 The d3/2state leads to the \m\ = 1 and 2
states and the d5/2 state leads to \m\ = 0, 1 and 2 states. Experimentally the
field ionization threshold fields of Fig. 7.11 are observed, when Na 17d
atoms are exposed to a pulse similar to Fig. 7.1(b) which rises in �0.5 jus
to the peak field shown on the horizontal axis. As expected from Fig. 7.10
there are only \m\ = 1 and 2 thresholds from the 17d3/2 state and \m\ = 0, 1,
and 2 thresholds from the 17d5/2 state. Had the Na nd fine structure intervals
been normal, the nd3/2states would only correlate to the \m\ = 0and 1 states
and the nd5/2 states to the \m\ =1 and 2states. Thefinestructure intervals of
the alkali atoms often fall inthe 1-10 MHzrange, in which case the transition
between spin orbit and uncoupled states can be made either diabatically or
adiabatically. Jeys etal.16have observed the transition from an adiabatic to a
diabatic passage from the coupled fine structure states to the uncoupled
states. With a pulsed laser, they excited Na atoms from the 3p1/2 state to the
34d3/2 state with a polarized light, which leads to 25% |ra;| = 1/2 atoms and

Spin orbit effects 117 (b) (c) uu rr QC D O z o 20 15 10 5 n -(a) 17d5/2 ^ / ^

^ ^ v ^ l i i 4.5 5.0 5.5 IONIZING FIELD (kV/cm) Fig.7.11 (a) Experimental
tracesofthe ioncurrent vs peak ionization voltagefor the 17d^/2 and 17d5/2
states. The approximate locations of the \m\ = 0, 1, and 2 thresholds are
indicated byarrows, (b),(c),(d)Oscilloscope tracesofionsignalsatdifferent
peak ionizing fields. In each case the center time marker corresponds to the
peak of the ionizing, high voltage pulse. The horizontal scaleis 200
ns/division. (b)m = 0 ionpulse,peak field = 4.58 kV/cm. (c)m = 0 followed
by \m\ = 1 ionpulse, peak field = 4.98kV/cm. (d) Overlapping m =0and \m\ =
1 ion pulses followed by \m\ =2 ionpulse, peak field = 5.27 kV/cm (from ref.
3). 75% \ntj\ = 3/2 atoms. After 100ns they applied a linearly rising field
ramp which rose to 0.4 V/cm at slew rates from 0.1 to 50 V/cm jus, a range
sufficient to encompass both adiabatic and diabatic passage. Afterwards a
fixed ramp of up to 800V/cm was applied which allowed the field ionization
signals of \m\ = 0,1,and 2 states to be resolved. As shown by Fig. 7.10 a
purely adiabatic passage yields25% \m\ = 1 and 75% \m\ = 2 atoms. A purely
diabatic passage yields 10% m = 0,30% \m\ = 1and 60% \m\ = 2 atoms.
When the slew rate was varied, the results shown in Fig. 7.12 were obtained.
At slew rates less than 0.5 V/cmJUS the passage is purely adiabatic and at
rates in excess of 10 V/cm JUS it is purely diabatic. These observations are
in reasonable agreement with estimates based on Eq. (7.6). The Na 34d fine
structure interval is 2.5 MHz16 and its tensor polarizability is 350
MHz/(V/cm)2.17 Eq. (7.6) is satisfied for E = 0.1 V/cm, and reaching this
field in a time of (400/2TT) ns, a slew rate of �1.5 V/cm jus should be the
approximate borderline between adiabatic and diabatic passage, and it is.
Making the transition from the coupled fine structure states diabatically has
been used in quantum beat experiments by Leuchs and Walther2 and Jeys et
al.18 We now consider the effect of the spin orbit coupling on the avoided
level crossings in high fields. With spin it is the projection on the z axis of
the total angular momentum, rrij = m + ms, which is conserved, not ms and m
separately. The spin orbit couplings between Stark states may be computed
just as we computed the quantum defect couplings. In essence the spin orbit
splittings are important if the � states involved have large enough spin orbit
splittings to produce noticeable differences in the quantum defects of
different; states of the

118 Rydberg atoms 100 Fig. 7.12 Ratio of the signal resulting from ionization
of \m\ = 2states (upper curve), and ionization of m = 0 states (lower curve) to
the total ionization signal as a function of the slewrate from
lowtointermediate fields following excitation ofthe 34d3/2 stateviathe Sp^
state with o polarization (from ref. 16). same �. For example in Na, the
pfinestructure interval produces a difference of 10~3in the quantum defects
of the np states. Thisisa factor of 15 smaller than the nd quantum defect. As a
result the avoided crossing of an m = 0, ms = 1/2 state with an m = 1, ms =
�1/2 state is negligibly small. Avoided crossings in which Am # 0are
traversed diabatically, and the spin maybe safely ignored, asitwas in the
earlier discussion of spinless atoms. In Li and H, which have smaller fine
structure intervals than Na, the effect of thefinestructure inhighfieldcan alsobe
ignored. On the other hand, in K the spin orbit splitting of the p states is large
enough that the \m\ = 0 and 1 levels are coupled strongly enough by the spin
orbit interaction that avoided level crossings are no longer traversed purely
adiabatically. Traversals of level crossings which are neither purely
adiabatic nor purely diabatic lead to multiple thresholdfields.As aresult,
states of \m\ = 0 and 1 of the same energy exhibit similar multiple threshold
fields.19 However, the optically accessible \m\ = 2 states exhibit single
adiabatic thresholds as expected from the small spin orbit splitting and large
quantum defect of the nd states. In Rb and Cs the npfinestructure is
alsosignificant, andmultiple thresholds areobserved for all

References 119 the optically accessible states, making field ionization a

much less selective detection technique.20 References 1. F. B. Dunning and
R. F. Stebbings, in Rydberg States ofAtoms and Molecules, eds. R. F.
Stebbings and F. B. Dunning (Cambridge University Press, Cambridge,
1983). 2. G. Leuchs and H. Walther, Z. Phys. A 293,93 (1979). 3. T. F.
Gallagher, L. M. Humphrey, W. E. Cooke, R. M. Hill, and S. A. Edelstein,
Phys. Rev. i4 16, 1098(1977). 4. T. H. Jeys, G. W. Foltz, K. A. Smith, E. J.
Beiting, F. G. Kellert, F. B. Dunning, and R.F. Stebbings, Phys. Rev. Lett. 44,
390 (1980). 5. D. S. Bailey, J. R. Hiskes, and A. C. Riviere, Nucl.Fusion 5,
41 (1965). 6. D. A. Park, Z.Phys. 159, 155 (1960). 7. W. E. Cooke and T. F.
Gallagher, Phys. Rev. A 21,580 (1980). 8. J. L. Dexter and T. F. Gallagher,
Phys. Rev. A 35,1934 (1987). 9. W. E. Cooke and T. F. Gallagher, Phys.Rev.
A 17, 1226 (1978). 10. J. R. Rubbmark, M. M. Kash, M. G. Littman, and D.
Kleppner, Phys.Rev.A 23,3107(1981). 11. J. L. Vialle and H. T. Duong, /.
Phys. B 12, 1407 (1979). 12. J. H. M. Neijzen and A. Donszelmann, /.
Phys.B 15, L87 (1982). 13. R. G. Rolfes, D. B. Smith, and K. B. MacAdam,
/. Phys. B 16,L533 (1983). 14. W. van de Water, D. R. Mariani, and P. M.
Koch, Phys.Rev. A 30, 2399 (1984). 15. G. B. McMillian, T. H. Jeys, K. A.
Smith, F. B. Dunning, and R. F. Stebbings,/. Phys.B 15, 2131 (1982). 16. T.
H. Jeys, G. B. McMillian, K. A. Smith, F. B. Dunning, and R. F. Stebbings,
Phys.Rev. A 26, 335 (1982). 17. T. F. Gallagher, L.M. Humphrey, R. M. Hill,
W. E. Cooke, and S. A. Edelstein, Phys.Rev. A 15,1937 (1977). 18. T. H.
Jeys, K. A. Smith, F. B. Dunning, and R. F. Stebbings, Phys.Rev. A 23,3065
(1981). 19. T. F. Gallagher and W. E. Cooke, Phys.Rev. A 19, 694 (1979).
20. T. F. Gallagher, B. E. Perry, K. A. Safinya, and W. Sandner, Phys.Rev.A
24, 3249 (1981).

8 Photoexcitation in electric fields Hydrogenic spectra A good starting point

isphotoexcitation from the ground state of H. The problem naturally divides
itself into two regimes: below the energy of classical ionization limit, where
the states are for all practical purposes stable against ionization, and above it
where the spectrum is continuous. As an example, we consider first the
excitation of the n = 15 Stark states from the ground state in a field too low to
cause significant ionization of n = 15states. From Chapter 6 we know the
energies of the Stark states, and we now wish to calculate the relative
intensities of the transitions to these levels. One approach is to calculate
them in parabolic coordinates. This approach is an efficient way to proceed
for the excitation of H; however, it is not easily generalized to other atoms.
Another, which we adopt here, is to express the n = 15nn1n2m Stark states in
terms of their n�m components using Eqs. (6.18) or (6.19) and express the
transition dipole moments in terms of the more familiar spherical n�m
states.1'2 In the excitation of the Stark states of principal quantum number n
from the ground state only p state components are accessible via dipole
transitions, so the relative intensities for light polarized parallel and
perpendicular to the static field, it and a polarizations, are proportional to the
squared transformation coefficients \(nn1n2m\n(m)\2 from the nn1n2m
parabolic states to the Mm states for � = 1and m = 0 and 1. In Fig. 8.1 we
show the relative intensities by means of the squared transformation
coefficients \(15n1n2m\15pm)\2 for m = 0and 1.When the exciting light is o
polarized, the distribution rises smoothly from both sides of the Stark
manifold. On the other hand, for itpolarization, the distribution drops
abruptly at both edges of the manifold. When the manifolds of adjacent n
overlap, the beginning of a new n manifold is much more apparent with it
polarized light. We can explicitly write out the statements above in the
following way. We denote the oscillator strength from the n'l'm! state to the
nriin2m state by fnmm,n'i'm'' From the Is state to the nnxn2m state it is given
by fnnim,iso = \{nnxn2m \npm)\2fnpmAs0 (8.1) where/npml50 = 2a>|
(npm|rm|ls0)|2, rm =zfor itpolarization and (x � iy)/2for a � polarization,
and co isthe photon energy. Since n2 is redundant, we do not use itin the
oscillator strength. Of the two factors in Eq. (8.1) the first varies within an n
manifold as shown by Eq. (6.19) and Fig. 8.1, and the second varies as n~3.
120

Hydrogenic spectra 121 0.3 ft 0.2c^ 0.1 if) (a) i . . 1 CNJ A 0.0 L 0 2 4 6 8
10 12 14 1 0.3 0.2c c" 0.1 (t>) ll ll0.0L 0 2 4 6 8 10 12 14 Fig. 8.1 Squared
transformation coefficients from the n = 15 parabolic 15 n\n2m states to
spherical 15p states for (a) m = 0and (b) m = 1.Note that the 15pstate
isconcentrated in the edges of the m = 0Stark manifold but at the center of the
m = 1 manifold. The procedure just outlined is perfectly adequate in the
regime in which the Stark states are all discrete. However, it breaks down
above the classical limit for ionization, when some of the Stark states
become continuous and n and n2 are ill defined. In this region the simplest
approach is to treat the problem as excitation to continua. We take advantage
of the fact that n1 is a good quantum number, compute the oscillator strengths
to the nx continua at energy W, and add these oscillator strengths to find the
total oscillator strength. Explicitly, the analogue to Eq. (8.1) is3 d f � = 2a*\
(Wn1m\rm\ls0)\2. (8.2) Summing over nt gives the total oscillator strength
,1*0 _ V 4/Wmm?isQ. (Q dw Zu To calculate the oscillator strengths of Eq.
(8.2) we transform the Wnxm continua to Wtm continua and calculate the
excitation in spherical coordinates.

122 Rydberg atoms First, we extend the transformation of Eq. (8.1) between
bound spherical and parabolic states into the continuum. How to go about
this was pointed out by Fano4 and Harmin.5 We are interested in the E ^ 0
Stark state wavefunctions near the origin, where the applied field is
negligible compared to the coulomb field, i.e. for r � W\fE. In this region
the wavefunctions for an E �^ 0 and an E = 0 parabolic state with the same
energy and separation parameters are functionally identical, and we may
replace the wavefunction for the parabolic E ^ 0state byits zero field
analogue with the same energy and separation parameters Z1 and Z2.
Harmin2 has shown that for continuum waves normalized per unit energy, in
zero field (Wnim\Wem)= (-lyPUZ, - Z2), (8.4) where P�m(Z1 � Z2) is a
normalized associated Legendre polynomial. Recall from chapter 6that for
any W, E, and ra, fixingthe value of nx fixes also the value ofZv In the small r
region, the only difference between the E = 0 and E ^ 0 wavefunctions with
the same separation parameters is in the normalization. That is for small r
\Wnim)E = ; (8.5) Here C� x is the density of states at the origin defined by
Eq. (6.33), and C�o is its zero field analogue. Using Eqs. (8.4) and (8.5)
aswell as the dipole selection rule, implying that � = 1 in the final state, we
can write the oscillator strength of Eq. (8.2) as d / � = 2a>\
(Wpm\rm\ls0)\2P2lm(Zi " Z2) -gfC Q 2 ( 7 7 ^ C Plm(Zl "Z2)o" ( } In this
form it is evident that the oscillator strength is simply the zero field oscillator
strength multiplied by P2m(Z1 � Z2), which gives the amount of p character
as nx or Zx is changed, and C�JC� l0 which gives the density of the nx
continuum at the origin. The zero field cross section and C^o are very slowly
varying functions of energy, so any structure is due to either C^ or P\m{Zx �
Z2). As shown by Fig. 6.7, far below the classical limit for ionization, C^ is
composed of delta functions corresponding to Stark states of good n2, and the
spectrum can be computed using the approach of Eq. (8.1). Near the classical
limit C� is rapidly varying, reflecting the presence of broader resonances,
the Stark states which ionize by tunneling. Above the classical limit C^J is
relatively smooth. In Fig. 8.2 we show the computed oscillator strength
distributions from the ground

Hydrogenic spectra 123 W(10"3a.u.) Fig. 8.2 nx continuum oscillator

strengths dfWnimls0/dW (shown as d//dW) and bound oscillator
strengths/nniml50 (shown as f) for transitionsfrom theground statetowards
n^m states in the H atom in the presence of the field of strength E = 1.5 x
10~5 au with it polarization (m = 0) and for energies smaller than the
ionization potential of the unperturbed atom. Vb is the critical energy in the
saddle point model; the parabolic critical energies Vb for the different nx
values are indicated by numbered arrows, the numbers givingthevalues of nx.
Continuum oscillator strengthsdfWnimls0/dW (lowerfigure)exhibit
resonances ofvery different widths. Structures associated with quasi-discrete
upper states, having a negligible width, but corresponding to a large value of
df/dW] are marked by a line. The resonances can be labelled bythe quantum
numbers n. nly and m. Byintegration of d//dWover the Lorentzian profile ofa
resonance, a totalvalue of the oscillator strength/ can be defined. The total
oscillator strengths are presented in the upper figure; a line connects the
different total oscillator strengths corresponding to different values ofnx, but
the same value of the principal quantum number n =l (-�-#-),n = 8 ( ), n = 9
( ), n = 10 (-�-�-), n = 11 (-O-O-). Each maximum in a curve dfWnimls0
/dW is related to awell defined/nniml50 with thesame nx and mvalues (from
ref. 3). state with JI polarization for n = 7 to n = 11 states in a field of 1.5 x
10 5 au (7.6 X 104 V/cm). Fig. 8.2 is analogous to Fig. 8.1(a). For n = 7to n
= 9the spectra are similar to the spectra of Fig. 8.1(a). The lower 0 < rii < 3
Stark states of n = 10 lie near the classical limits for their nx channels and
mark the onset of continuous absorption in these channels. However, the
higher n = 10 Stark levels of nx > 5still appear assharp lines. Of the n =
11states, only the highest energy, blue, Stark states are recognizable as
resonances; the lower Stark states have become part of the continuum
absorption. A point which is important to note inFig. 8.2 isthat absorption
corresponding to the center of the

124 Rydberg atoms w(10-3a.u.) Fig. 8.3 The same asinFig. 8.2 but with
opolarization, m = 1 (from ref. 3). manifold, where nx = n2 = nil and Zx = Z2
= 1/2, isalwaysweak whether the final states are discrete or a continuum. In
Fig. 8.3 we show the oscillator strengths for excitation from the H ground
state with a polarization, to excite final m = �1 states. The oscillator
strength distribution for 7 < n < 9 resembles the oscillator strength
distribution of Fig. 8.1(b). At n = 10 the onset of the continuous spectrum is
observed in the lower Stark states but not the upper Stark states, as in Fig.
8.2. The fundamental difference between Figs. 8.2 and 8.3 is that in Fig. 8.3
there is appreciable oscillator strength from the ground state tothe center of
the Stark manifold, not to the edges. Above the zero field limit, where one
would not expect to see much, if any, structure in the photoionization cross
section, there exist regular oscillations, as first observed by Freeman et al. in
the photoexcitation of Rb from its ground state.6'7 They excited a beam of Rb
atoms with a frequency doubled pulsed dye laser to obtain the spectra shown
in Fig. 8.4. They observed oscillations with n polarized light but no
oscillations with a polarized light. These oscillations are not peculiar to Rb
but exist in H as well and can be described by computing the oscillator
strength using Eq. (8.6). Since the ep oscillator strength is so slowly varying
as to be effectively constant, the modulation in the signal apparently
originates in the density of states, CJJi and P\m(Zl � Z2). The density of
states near the origin, C�, for a specific n1? is only nonvanishing over an
energy range corresponding to a range of values of Zx and Z2. While the
Hydrogenic spectra 125 ZERO FIELD IONIZATION .GAIN^O^LGAIN^J ^ j
c j 10 GAINxiO GAINx5 I GAIN x 2 i OFFSET^-22 OFFSET-a
OFFSET�-3 3005 Fig. 8.4 Relative ground state photoionization cross
section as a function of laser wavelength in Rb in the presence of a 4335
V/cm field. Note the relative gain and offset settings. For the light
polarization parallel to the electric field (lower trace), field dependent
resonance structure extends beyond the zero field limit. No structure is
observed for thecase of light polarized perpendicular to the field (upper
trace) (from ref. 6). statement istrue for allra,it isparticularly easy to compute
the energy range for m = 1, in which case the potentials V(|) and V(rj), from
Eq. (6.23), have the forms ( ^ ^ ) (8.7a) and From Eq. (8.7a) it is apparent
that for energies near the ionization limit the electron is excluded from the
origin in the � motion if Zx < 0.Similarly ifZ2 < 0 the electron is excluded
from the origin in therjmotion. Therefore Zx =0 and Z2 = 0 represent the
limits of the significant values of <XX �Stated in terms of Zx only, C\
isnonzero over the range 0 < Zx < 1.For a fixed nx, Zx decreases with
increasing energy. We may estimate the energies at which Zx =0and 1 using
the WKB approximation. Examining Eqs. (6.22a) and (6.23a) we can see that
if we set r(H?rM where �m is the classical outer turning point, and evaluate
it forZx � 0and 1 we can determine the upper and lower energy bound of
Clm foreach value of nx.3'8 In Fig. 8.5 we show the computed values ofC�
for nx = 13 m = 0 and nx = 12 m = 1.3 Both are nonzero over similar energy
ranges, as expected. However, CPX3 has sharper edges than C\2. Also
shown in Fig. 8.5 are the oscillator strengths d/Wmm,i5o/dW which are
proportional to C^PJm(Zx - Z2). Due to the fact that Pio(Zi ~ Zi) vanishes
forZx - Z2 = 0and is peaked atZx - Z2 = �1, the edges of Cni� contribute
peaks to the m = 0 cross section while thecenter contributes

WdW(10"�a.u.) Fig. 8.5Comparison between the energy dependences of the

partial density of states C� (�) and the continuum oscillator strengths
dfWnimXs0 /dW (shown asdf(nx ,m)/dW) ( ) in the photoabsorption
spectrum from the ground state of H inthe presence ofan external electric
field E = 1.5 x 10~5 au at energy greater than 0, the ionization potential of the
unperturbed atom, for nx � 13,m = 0 andnx = 12,\m\ = 1; Wc is theenergy
atwhichthe nx channel becomes continuous, corresponding to Zx ~ 1and Z2~
0. Wd is the energy at which Zx � 0 and Z2 � 1 (from ref. 3). -dL(n,,O)
_dL(O) 0.6 3 a.u.) "47 E o 18 +5 W(10"3a.u.) Fig. 8.6Energy dependences
ofthe continuum oscillator strengths dfWnimXs0 /dW(shown asdf(nx,m)/dW)
for \m\ = 0and 1 in the photoabsorption spectrum from the ground state of H
in the presence of an external electric field E = 1.5 x 10~5 au. The total
oscillator strengths dfWmls0/dW (shown as d/(0)/dW) are presented at the
top of thefigure;the depth of the modulation at the zerofieldionization
potential is equal to 18% and1%for m =0 and 1 respectively (from ref. 3).
nothing. In contrast, when multiplied by P\X{ZX � Z2) the center of Clm
contributes a single broad peak. When the oscillator strengths for all values
of nx are summed, the sharp peaks on the lowenergy sides of the m = 0rii
channels are still evident, while the broad peaks of the m = 1 channels add to
give a nearly structureless oscillator strength, as shown by Fig. 8.6.
Examining CPl3 and C\2 in Fig. 8.5 it is tempting to conclude that the
relatively sharp low energy edge of C^1 would, by itself, lead to observable
structure in the cross section. However, as

Hydrogenic spectra 127 pointed out by Luc-Koenig and Bachelier, the

modulation in the cross section would then be �3%, far less than observed
experimentally.3'9 The variations of both CPm and P\\{ZX � Z2) areneeded
for observable modulations inthe oscillator strengths. If we are only
interested in the frequency of the modulations in the vicinity of the zero field
limit wemay employ adifferent approach, used byFreeman etal.6'7 and
Rau10. They used the fact that the motion in the � direction is bound and
found the energy separation between successive eigenvalues. Specifically,
they used Eq. (8.8), the WKB quantization condition for the bound motion in
the � direction, and differentiated it to find the energy spacing between
states of adjacent nx or, equivalently, between the oscillations observed in
the cross sections. Differentiating Eq. (8.8) with respect to energy yields 4
dW Since we are interested in the result for Zx = 1near W = 0, we set W = 0
and Zx = 1 and choose �m = 2l\rE. Evaluating the integral by makingthe
substitution sin 6 = VE^/2 we find � = 3.70�3/4. (8.10) dnx In laboratory
units, dW/dnj = 22.5 cm"1 when E = 4335 V/cm. In other words, the spacing
between the high energy edge of the continuous spectrum of Clm for adjacent
values of nx scalesas �"3/4in thevicinityof theionization limit, W = 0. Eq.
(8.10) correctly describes the frequency of the oscillations in the cross
section, as shown in Fig. 8.4. Another approach to calculating the spectra
such as the one of Fig. 8.4 was suggested by Reinhardt,11 who proposed
using a wave packet approach. The essential idea is that the laser light
creates a wave packet at the origin (more precisely in the volume of the
ground state) which propagates radially outward from the origin. If the wave
packet encounters a potential barrier, such as the Stark potential on the
upfield side,it is reflected. If thewavepacket returns to the origin during the
laser pulse astanding waveis created, leading to the modulation in the cross
section. How strong the modulation is depends upon how large a fraction of
the outgoing wave packet returns to the origin. This picture clarifies why
there is only modulation with JI polarization. With n polarization half the
electrons are preferentially ejected in the upfield, +z, direction and are
reflected back to the origin. With a polarization no electrons are ejected in
the upfield direction and none are reflected back to the origin. A related
approach to the calculation of the strong field mixing spectrum has been
pursued by Gao etal.12They start the calculation in the sameway Reinhardt
does, by computing the outgoing quantum mechanical wavefunction of the
photoelectron produced at the origin by photoabsorption. At a distance �50
a0

128 Rydberg atoms from the nucleus the outgoing wave fronts are converted
to classical trajectories, perpendicular to the wave fronts. The classical
trajectories are then propagated. Some of the trajectories in the upfield
direction are reflected back to the origin, and, before they approach the
origin, they are converted back to quantum mechanical wave fronts to
recreate the quantum mechanical wavefunction. The constructive and
destructive interference of the returning wavefunction with the outgoing
wavefunction leads to the modulations in the strong field mixing spectrum.
This approach leads to an excellent representation of the depth of the
modulation as well as the locations of the maxima and minima of strong field
mixing resonances in Na.12 These are two quantities which are not specified
by the WKB approach of Eqs. (8.9) and (8.10). The oscillations in the
photoionization cross section are not restricted to starting from an sstate.
They can be observed from other �initial states aswell, as demonstrated by
Sandner et al.,13 who observed the photoexcitation from the excited Ba
6s6pstates in electricfieldsusing the twolaser excitation scheme Ba6s2 ��
6s6p �> 6s�S, 6sed. They used the four choices of linear polarization JZ-
JZ, JZ-O, O-JZ, and o-o., and the O-JZ and JI-JI spectra are shown in Fig.
8.7. The features at W = 42,117 and 41,841 cm"1 are doubly excited states,
and the ionization limit is at 42,035 cm"1. For JZ-JC polarization strong
oscillations are observed, for o-o polarization weak oscillations are
observed, and for o-n and JZ-Opolarizations no oscillations are observed.
The oscillations in the JZ-JZ and o-o spectra come from the excitation to the
ed m = 0 continuum (The oscillator strength to the s continuum is much
weaker and can be ignored.). For the n-n spectrum of Fig. 8.7(b) the analogue
to Eq. (8.6) is given by = d/wd0,6pQp2 _ C� dW v J C% The only
functional difference between Eqs. (8.6) and (8.11)is in the Legendre
polynomial. PW{ZX � Z2) is even more sharply peaked at Zt � Z2 = �1
than is ^io(Zi � Z2). Consequently, in the oscillator strength the low energy
side of C�x is more strongly emphasized in Eq. (8.11) than in Eq. (8.6).
Accordingly, the JZ-JZ oscillations in Fig. 8.7(b) are more apparent than
those in the JZ spectrum of Fig. 8.4. The expressions analogous to Eq. (8.11)
for the O-JZ and JZ-O polarizations contain P\X(ZX � Z2), while for o-o
polarization it contains both P^oi^i ~ Z2) and P22(ZX � Z2), the latter being
dominant. Neither P\\{ZX � Z2) nor ^22(Zi � Z2) is strongly peaked at Zx
� Z2 = 1, and there are no visible oscillations in the JZ-Oor O-JZ spectra
and only asmall oscillatory component in the o-o spectrum. In the H atom it is
not possible to excite Rydberg states in a strong electric field from an angular
momentum state other than the Is state. All other states are converted to Stark
states by the field, leading to a pronounced asymmetry in the excitation of red
and blue Rydberg Stark states from them. The asymmetry is easily understood
by considering excitation from the n = 2, m = 0Stark states. In

Hydrogenic spectra 129 4 c I3 15 .2t 2 42,117 42f035 41,930 41,841 v

(cm"1) (b) Ba 4.80 kV/cm 7T-7T I I I I Aik i I | f if142,117 42,035 41,930
41,871 v (cm"1) Fig. 8.7 Experimental photoabsorption signal in Ba at an
electric field strenth of 4.80 kV/cm, obtained by resonant two photon
absorption via the intermediate 6s6p1P1 state with (a) O-JT, and (b) JT-JZ,
polarization (from ref. 13).

130 Rydberg atoms the red state the electron ison the �zside of the proton
and the blue state it is on the +z side of the proton. It istherefore hardly
surprising that photoexcitation of red states is to red states and blue statesis
to blue states. A straightforward way to make this notion quantitative is to
express the Hm =0,n =2 \ipnmm) Stark states in terms of the n = 2 \nlm)
states. Explicitly, IV200)= kred) = ^ (|2s0> + |2pO�, (8.12a) and IVW =
IVbiue)= ^ (|2sO>- |2p0�. (8.12b) In calculating photoexcitation cross
sections from either of these two parabolic states we must coherently add the
transition amplitudes from the s and p partsof the n = 2state. Consider, for
example, the excitation of the m =0Stark statesof principal quantum number
natfieldswellbelowthefieldsrequired for ionization. The relative strengths of
the transitions are given by expressions analogous to Eqs. (8.1) and (8.2).
Explicitly, the oscillator strength fnmo,2nlof�r excitation of the nn1n20 state
from the n = 2, m = 0 Stark states is given by (nsO\z\2vO) I (m1n20\nd0)
<m!0|z|2p0>} , (8.13) where the + and � signs refer to the red (n[ =0) and
blue (n[ = 1) n = 2, m =0 states respectively. Using the associated Legendre
polynomial form of Eq. (6.19) for the transformation coefficients, Eq. (8.13)
may be written as As n -> oo, (?zpO|z|2sO>, (nsO|z|2pO), and <ndO|z|2pO),
are given by 3.83 x n~3/2, 1.11 x n~3/2, and 3.95 x n~3/2, respectively.14
Thus, the dominant interference among the three terms of Eq. (8.14) is
between the excitations to the np andnd components. Both P10 {{tii� n2)/n)
and P2o ((wi ~ ^2)/^) have their maximum magnitudes at (nx � n2)ln = � 1,
and consequently their interference is most pronounced inthe extreme blue,rii
~ n, or extreme red, n1 ~ 0, statesofthe Stark manifold. Since P2o ((ni ~ n
i)lri) is positive for (nx � n2)ln = � 1 but P10 ((�! � n2)lri) changes sign
with(nx - n2)/n, the excitation amplitudes to the blue or red final state add
constructively or destructively depending on whether the initial state is a
blue or red n = 2state.

Hydrogenic spectra 131 o c (a) 1 . . . 1 1 1 1 i . . . 8 10 12 14 CM O in (b) . .

i I I I l . . . 1 8 10 12 14 Fig. 8.8 Calculated oscillator strengths to the n =
15,m = 0 Stark states from the (a) red n =2, nx = 0, m =0 Stark state (b) blue
n =2, n1 = l,m =0Stark state. In Fig. 8.8 we show a graph of the oscillator
strengths from the red and blue n = 2 Stark states to the n = 15 Stark states,
i.e. /i5nio,2�o> assuming that co has its zero field value of 0.123. As shown
by Fig. 8.8, from the red n = 2 state predominantly the red n = 15 states are
excited, while from the blue n = 2 state predominantly the blue n = 15 states
are excited. The result of exciting from the red instead of the blue n = 2 Stark
state is simply to reverse the asymmetry of the excitation of the Stark
manifold.
132 Rydberg atoms _J SIG f cr u h~ O UJ _J UJ 0 (a) (b) ***** Q ZER O
FIE I I.R t ZER O FIEL D H ^ ^ ^ ^ 2n" Ti 11 it hJL.ltLj..i^..J.L, iJii 5 c 11 1
rrm i| JILl 1, ill Tn xL _ : d TI Ji ^100 ^200 : <3O6 -400 ENERGY (cm-1)
Fig. 8.9 Observed spectrum from the red n = 2, nx =0, m =0 Stark state in
afieldof 5714 V/cmwith (a) it and (b) apolarization ofthe second laser. Note
the sharp onset ofthe continuous spectrum (from ref.15). In stronger fields,
the difference between starting from the red and blue states becomes more
pronounced. Since the red states are more easily ionized than the blue states,
we expect to observe the onset of a continuous spectrum sooner and more
abruptly when exciting from the red n = 2 state than from the blue n = 2 state.
This point has been clearly shown by Rottke and Welge15 and Glab etal.16
Rottke and Welge excited H atoms in a beam first to one of the four n = 2
Stark states with a vacuum ultraviolet laser at 1216A which wasfixedin
frequency, and then used a second tunable laser to excite the H atoms from
the selected n = 2 Stark state to the region of the zero field limit. The
experiment was done in the presence of a static electric field of 5-10 kV/cm
and the electrons resulting from the excitation and subsequent ionization of
Rydberg states were detected as the wavelength of the second laser was
scanned. They detected atoms which ionized with rates of 5 x 105 s"1 or
faster. In Figs. 8.9 and 8.10 we show the spectra obtained using the red and
blue n = 2, m = 0 Stark states as intermediate states. Fig. 8.9(a) shows the
spectrum obtained by sweeping the second, Jt polarized, laser using the red n
= 2 Stark state as the intermediate state. As shown, the spectrum is dominated
by the continuous excitation of the red Stark states. The

Hydrogenic spectra 133 T Tn m t i TTTTTTT S 2 > S- 2- I I I II I "� �� z

-: rTTTI III I I M 5 1 g lk Jill I..JJ �100 0 -100 -200 -300 -400 ENERGY
(cm"1) Fig. 8.10 Observed spectrum from the blue H� = 2,n1 = l,m = 0
Stark state in a field of 5714 V/cm with (a) it polarization and (b) o
polarization of the second laser. Note the progression from the extreme blue
Stark states to the strong field mixing resonances (from ref. 15). most
prominent features are the relatively low nx Stark states of n = 18 which lie
just above their classical ionization limits. Fig. 8.10(a) shows the analogous
excitation spectrum obtained with Jt polarization using the blue n = 2 Stark
state as the intermediate state. There are several marked differences between
Figs. 8.9(a) and 8.10(a). First, there is no sharp onset of continuous
excitation in Fig. 8.10(a), rather agradually growing continuum excitation.
Second, there are many more sharp states excited, and they are mostly to the
blue sides of the Stark manifolds. From Fig. 8.8 wecould reasonably expect
predominantly blue states to be excited. The fact that there are far more sharp
blue states than red states is due to the fact that their ionization rates do not
increase as rapidly with field as do those of the red states, as shown by Fig.
6.8. Finally, we note that evolution of the extreme blue Stark states into the
strongfieldmixing resonances isclearly evident in Fig. 8.10(a). The
resonances are spaced as E3/4, as described by Eq. (8.10), but the
modulation is in this case deeper than it is when starting from an initial state
which is spherical. The modulation is deeper for the same reason that Fig.
8.8 is

134 \f) LLJ h00 LL. o 4SI T LJ Q TI A cc CL d z o 100 00 00 o o cr 0.1 0.08

0.06 0.04 0.02 0 0.2 0.15 0.1 0.05 0 i \ i Z, = 1 ^ n i = 2 \ \ \ l \ \ \ \ ^ II i 0 100
Rydberg i ^ ^ ^ ^ 5 6 ^** l 200 ENERGY atoms i i i i / / / 1 � * i i 300 400
fern"1] 11 ' llz^o _ (a) \ | (b) \ � .vl 500 Fig. 8.11(a) Calculated density of m
= 0 states CPni abovethezerofieldionization threshold forfinalstateswith
quantum numbers nx =25,26at E = 5714 V/cm. Positions Zx =0 and1 are
indicated by the arrows, (b)Calculated oscillator strengthsfor excitation from
then = 2 parabolic states 210(blue) and 200 (red) into the channel with
quantum number nx =26 at 5714 V/cm in the energy region W> 0. Curve I,
d/^lO2io/dW; curve II, dfWniO2oo/dW (fromref. 15). not symmetric in the
red and blue final states. The oscillator stength to the nx continuum from the n
= 2, m = 0 red and blue Stark states is given by d / ' ^ - Z2)(�p0|z|2s0) �
(PooiZ, - Z2)<�s0|z|2p0> + P20(Z1 - Z2)<�d0|z|2p0�]2, (8.15) where the
+ and- sign refer toexcitation from the red (n[ =0)andblue (n[ = 1) n = 2 state
respectively. Inspecting Eq. (8.15) wecan see that, apart from constant
factors, it issimplya continuousversion of Eq. (8.14),orFig. 8.8, multiplied by
c�,.Thispoint is made explicitly in Fig. 8.11 in which CPm and the
oscillator strength are plotted. For an initial blue n = 2 state Fig. 8.8 peaks at
nx = n, or Zx = 1, corresponding to the sharp low energy edge of C^ (curve I
of Fig. 8.11(b)). In contrast, for the red

Nonhydrogenic spectra 135 initial state Fig. 8.8 peaks at rt\ � 0, or Z x = 0,

corresponding to the more gradually sloped high energy side of C^ (curve II
of Fig. 8.11(b)). When the oscillators strengths of many nx channels are
added, the oscillations in the spectrum from the n = 2 blue state are stronger
than those in the spectrum from the Is state, while those in the spectrum from
the red n = 2 state almost vanish. Realizing that the strong field mixing
resonances evolve from the extreme blue Stark states we can write their
energies to first order in the electric field as W=-JL + ^ , (8.16) and the E3/4
spacing of the resonances at W=0 follows immediately from Eq. (8.16).
Nonhydrogenic spectra The photoexcitation of nonhydrogenic atoms is in
many ways similar to the photoexcitation of H. For example, the
strongfieldmixingresonances observed in Rb, Ba, and Na are well described
by a hydrogenic theory.6'713 However, all features of the photoexcitation
spectra of nonhydrogenic atoms are not equally well described by a
hydrogenic theory, and we now describe the deviations. It is convenient to
consider three spectral regions, below the classical ionization limit, above
the classical ionization limit but below the zero field limit, and above the
zero field limit. Below the classical ionization limit all the states are
effectively bound, and the probability of exciting aStark state from
alowlyingntm state maybe computed in the same way we computed it for H.
Specifically, we project the Stark state of interest onto the zerofield�states
and add the excitation amplitudes from the low lying state. Unfortunately,
wecannot use the transformation from the Stark states to the nim states given
by Eq. (6.19). An excellent illustration of the deviation of the transformation
coefficients from the hydrogenic values is afforded by the excitation of the Li
m = 0states of n =15 from the 3s state, shown inFig. 8.12. The spectra
showninFig. 8.12were obtained by Zimmerman et al.17 by exciting Li atoms
in a beam from the ground state via the 2p state to the 3sstate using two
pulsed dye lasers. A third dye laser, polarized parallel to the field, drove the
3s �� Rydberg transition and was scanned in frequency to produce each of
the spectra of Fig. 8.12. The Rydberg atoms were detected by applying a
field pulse subsequent to excitation to field ionize them. Since only the p m =
0 components of the Stark states are accessible, we might expect a field
independent distribution of oscillator strength among the Stark states,
asimplied by Fig. 8.1. As shown byFig. 8.12, infieldsup to �1500 V/cm all
the Stark states are excited, but not with the field independent hydrogenic m =
0

136 Rydberg atoms 1 1 1 1 1 1 i 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 440 450 -E 460 ~z
1000 2000 3000 4000 5000 6000 FIELD (V/cm) Fig. 8.12Spectrum of the
Lim = 0 Stark states from aninitially excited 3sstate. The line intensities are
proportional to the amount of 15p m =0 character in each Stark state (from
ref. 17). pattern shown in Fig. 8.1(a), rather with a pattern more similar to
the m = 1 pattern of Fig. 8.1(b). Furthermore, the pattern is not field
independent; for E > 1500 V/cm the excitation of the members near the
edgesof the Stark manifold disappears, only to reappear at higher fields. In
this regime, where the levels are discrete, it is possible to calculate the
intensities of the transitions by matrix diagonalization, just as the energies are
calculated. It is simply a matter of computing the eigenvectors of the
Hamiltonian as well as its eigenvalues. For example, to calculate the
intensities in the spectra shown in Fig. 8.12 we calculate the np amplitude in
each of the Stark states and multiply it by the matrix element connecting the
3s state to the n'p state, JnmO,3sO ~ - (8.17) Often only one value of n' is
really significant. For example for the n = 15states of Fig. 8.12 only then' =
15term of Eq. (8.17) makes a large contribution. As shown by Zimmerman et
aL, this procedure allows the intensities of Fig. 8.12 to be predicted
accurately.17 As shown in Fig. 8.12, even the small Li+ ionic core,
producing an s state quantum defect of 0.3, radically alters the spectrum of Li
m = 0 atoms from the

Nonhydrogenic spectra Yil H m = 0spectrum. With larger quantum defects the

departures can be even more extreme, asshown by the energy level diagrams
of the Na m = 0and1states of Figs. 6.11 and 6.12. The energies ofthe Na m =
0levels bear no obvious relation to hydrogenic levelsfor E > l/3n5.
Furthermore, it is often difficult to pick out any pattern. Nonetheless, there
does exist a similarity to the hydrogenic spectrum, which is brought
outbythetechnique of scaled energy spectroscopy. The basic notion is as
follows. Classically, the Hamiltonian forthe H atom in the field Ez is H =
p2/2 - IIr + Ez, (8.18) wherep2 is the momentum of the electron. It obeys
classical scalinglaws, and ifwe make the replacements f = rEm p=pE-m
(8.19) E=l. then 6(p,f,l ) = H(p,r,E)/VE, (8.20) and the scaled energy W =
W/VJE. Eichmann et al.ls obtained spectra of Na at constant scaled energy W
= -2.5 by simultaneously scanning both the electric fieldand the wavelength
of the exciting laser so as to keep W = -2.5. This scaled energy lies slightly
below the classical ionization limit, where W = �2.0. They excited Na
atoms in abeam to either the 3p1/2 or3p3/2 state using a laser linearly
polarized in thefielddirection. With this polarization m =0and 13patomsare
produced. They then excited atoms from the3pj state to states of W = -2.5 by
scanning the frequency of a second laser while the field was scanned to keep
W � 2.5. Subsequent to the laser excitation afieldionization pulse was
applied to the atoms to ionize them and accelerate theresulting ions toward
the detector. The resulting spectrum, plotted vs laser frequency, is
abewildering mass oflines suggesting no obvious interpretation. However,
when it isFourier transformed to obtain the spectrum vsthe classical action 5,
it exhibits surprising regularity and is very similar to the Fourier transform
spectra computed for H. Fig. 8.13 illustrates this point. In Fig. 8.13(a) we
show the measured Fourier transform spectrum obtained with apolarization
ofthe second laser, so as to produce predominantly m = 2 final states.
Theexperimentally derived spectrum of Fig. 8.13(a) isvery similar to the
theoretical spectrum of Fig. 8.13(b) computed for the excitation H m = 2
states from the 2p m = 1 states (to observe such transitions is, ofcourse,
impossible). Since the Na m � 2 states arevery nearly hydrogenic, the
similarity of Figs. 8.13(a) and 8.13(b) demonstrates primarily the agreement
of theory and experiment in a known test case. InFig. 8.13(c) we show the
observed spectrum when the second laser is Jt polarized, so asto produce m
= 0 and 1finalstates. This experimental spectrum is similar toFig. 8.13(d), the
calculated spectrum for Am = 0transitions from the H 2pm = 1 state, with the
largest discrepancies occurring at the highest values of the action S.

138 Rydberg atoms in *-� ). un i U-l L_ o ,/|.J,,AJlii (a) (c) o CD Q. CO L.

CD O OL 8 10 0 (b) -V^V JlAft,ll 0 2 A 6 8 10 0 2 / . 6 8 Classical Action S
(a.u.) Fig. 8.13Power spectra of measured Naphotoexcitation cross
sectionsfrom the3p states vs the classical action S in atomic units: (a) a
polarization and (c) JI polarization; and calculated power spectra ofH (b)
from the 2pm = 1state with opolarization and (d) from the 2p m = 1 state with
n polarization. Allareforfixedscaled energy W= W^/E = �2.5 (fromref. 18).
Above the classical ionization limit the blue states, which are stable in H,
are degenerate with the red continua. In H they are not coupled, but in any
other atom they are. We have already described how this coupling leads to
ionization, which is a form of autoionization.19 Similarly, the coupling leads
to the same sort of interference in the excitation amplitudes that produces the
asymmetric Beutler-Fano profiles common in the excitation spectra of
autoionizing states.20 Ultimately the electron finds its way into the red
continuum, but it may reach the continuum by either direct excitation or
bypassing through the blue state, and the amplitudes for these two paths
interfere. Since the strength of the blue state excitation varies across the state
and the phase of the continuum changes by jt across the state, at some point
the two amplitudes are equal and of opposite sign, resulting in the
crosssection vanishing. Of course if there isasecond, noninteracting,
continuum, it contributes a constant cross section, and the total cross section
does not vanish at any energy.20 In Fig. 8.14 we show an example of an
asymmetric Beutler-Fano profile observed in am = 0state of Rb in afieldof
158 V/cm.21 The spectrum of Fig. 8.14 was obtained by Feneuille etal.21 by
exciting Rb atoms in a well collimated beam

Nonhydrogenic spectra 139 1GHz 1GHz (b) Fig. 8.14 (a) Photoionization
spectrum of the ground state of Rb in the presence of a 158 V/cm static field
for an excitation energy in the vicinity of 33,614 cm"1 and a light
polarization parallel to the field, (b) Example of a charcteristically
asymmetric profile. Dotsrepresent thebestfittoa Beutler-Fano profile.
Thisfithas beenobtained by assuming a linear variation of the ionziation
background vs the excitation energy (dotted line). Fabry-Perot fringes 1.3
GHz apart provide thefrequency scale (from ref. 21). with a single mode
pulsed dye laser, frequency doubled to provide uv pulses of 40 MHz
linewidth. The atoms are excited in the presence of a static field which
accelerates the Rb+ ions formed to a particle detector. As shown by Fig.
8.14, in most Stark spectra above the classical ionization limit there is never
one isolated resonance but, more often, an irregular jumble of them. For
example, in Fig. 8.15 we show the observed22 and calculated23 Na spectra
near the ionization limit in a field E = 3.59 kV/cm.22 The experimental
spectrum of Fig. 8.15(a) was obtained by Luk etal.22 by exciting a Na beam
with two simultaneous dye laser pulses from the 3s1/2 to 3p3/2 state and then
to the ionization limit. Both lasers were polarized parallel to the field, and
the ions

140 Rydberg atoms m=0 ((22,2) (23,0) (23,1) (23,2) (24,0) (24,1) (25,0)
(26,0) | I Na 32P3/2 E = 3.59 kV/cm 24500 Fig. 8.15 (a) Experimental
photionization spectrum from the Na 32P3/2 state in a field E = 3.59 kV/cm,
vs photon energy ho, within 0.01 eVof threshold (from ref. 21). Both the laser
populating the 3p3/2 state andthe second scanned laser arepolarized parallel
to the staticfield. Note labeling ofStark resonances {nx,n2) form = 0 and 1.
(b)Theoretical cross section, from thedensity of states theory. Both m = 0 and
1 final states are present (from ref. 23). resulting from photoexcitation were
collected as the wavelength of the second laser was scanned. The resulting
spectrum of Fig. 8.15(a) is a composite of m = 0 and m = 1final states since
the spin orbit splitting of the 3p states, 17cm"1, is far greater than the inverse
of the pulse duration of the lasers. Reproducing an entire experimental
spectrum such as Fig. 8.15(a) poses a formidable challenge to the theorist,
for the entire spectrum can not be interpreted using an isolated resonance
approach used to interpret Fig. 8.14. However, Harmin's density of states
approach23 has been very successful in reproducing experimental spectra, as
shown by Fig. 8.15(b). The density of states theory, a form of quantum defect
theory applied to the Stark effect, also accounts for, quite naturally, the
variation in the ionization rates, which is aquite general effect occurring in
allthree or more channel systems.24 In the density of states theory space is
divided into three regions. For r < r0 the effect of the external field may be
neglected, and for r > rc the potential from the

Nonhydrogenic spectra 141 ionic core is negligible. Thus only in the region
rc < r < r0 is the potential a pure coulomb potential, and in this and only this
region may we transform freely between the spherical and
parabolicwavefunctions. The noncoulomb potential at r< rcintroduces a
radialphaseshift of 6%intothe sphericalwavefunctions withthe result that they
are given by a linear combination of the regular and irregular coulomb
functions in the region rc< r < r0. In this region the wavefunction in spherical
coordinates is given by V = ' " ^ (cos dtrfir) + sin d(rg(r)), (8.21) where/(r)
and g(r) are the regular and irregular coulomb functions. Since the g function
is irregular at the origin, it is evident that the wavefunction cannot be
expressed in terms of parabolic functions regular at the origin. Harmin
expresses the parabolicwavefunction interms of a regular function of �and a
regular and an irregular function of rj. Explicitly,23 eim6> ) cos <$s+ xiiv)
sin <5S] y==, (8.22) where^.(rj) is the irregular solution of Eq. (6.22b) for
Xz(v) valid for r> rc. In the classically allowed region ^(rj) is oscillatory,
90� out of phase with XiivDSpherical wavefunctions of the form of Eq.
(8.21) may be projected onto the parabolic wavefunctions of the form given
by Eq. (8.22). There are two points to note regarding Eq. (8.22). First, the
phase ds is related to the zerofieldquantum defects bythe transformation
which transforms the spherical wavefunction of Eq. (8.21) into the parabolic
wavefunction of Eq. (8.22). Second, the wavefunctions of Eq. (8.22) are not
orthornormal due to the presence of both regular and irregular functions
%2(^) an d ^xAv)-- When the wavefunctions analogous to the one of Eq.
(8.22) are orthonormalized, and the spherical wavefunctions reexpressed in
terms of them, the transformation includes interference terms between the
orthonormalized final states. These terms, which vanish if all the quantum
defects are zero, lead to the interferences observed in the excitation spectra
of nonhydrogenic atoms. References 1. D. A. Park, Z. Phys. 159,155 (1960).
2. D. A. Harmin, inAtomic Excitation and Recombination in External Fields,
eds. M. H. Nayfeh and C. W. Clark (Gordon and Breach, New York, 1985).
3. E. Luc Koenig and A. Bachelier, /. Phys. B 13,1769 (1980). 4. U. Fano,
Phys.Rev. A 24, 619(1981). 5. D. A. Harmin, Phys.Rev. A 24, 2491 (1981).
6. R. R. Freeman, N. P. Economou, G. C. Bjorklund, and K. T. Lu, Phys.Rev.
Lett. 41,1463 (1978).

142 Rydberg atoms 7. R. R. Freeman and N. P. Economou, Phys.Rev. A 20,

2356 (1979). 8. D. ter Haar, Problems in Quantum Mechanics (Pion, London,
1975). 9. E. Luc-Koenig and A. Bachelier, Phys.Rev. Lett. 43,921 (1979).
10. A. R. P. Rau, /. Phys. B 12,L193 (1979). 11. W. P. Reinhardt, /. Phys. B
16,635 (1983). 12. J. Gao, J. B. Delos, and M. C. Baruch Phys.Rev. A 46,
1449 (1992). 13. W. Sandner, K. A. Safinya, and T. F. Gallagher, Phys.Rev.
A 23, 2448 (1981). 14. H. A. Bethe and E. A. Salpeter, Quantum Mechanics
of One and Two Electron Atoms (Academic Press, New York, 1957). 15. H.
Rottke and K. H. Welge, Phys. Rev.A 33, 301 (1986). 16. W. L. Glab, K. Ng,
D. Yao, and M. H. Nayfeh, Phys.Rev.A 31, 3677 (1985). 17. M. L.
Zimmerman, M. G. Littman, M. M. Kash, and D. Kleppner, Phys.Rev. A 20,
2251 (1979). 18. U. Eichmann, K. Richter, D. Wintgen, and W. Sandner,
Phys.Rev. Lett. 61, 2438 (1988). 19. M. G. Littman, M. M. Kash, and D.
Kleppner, Phys.Rev. Lett. 41,103 (1978). 20. U. Fano, Phys.Rev. 124, 1866
(1961). 21. S. Feneuille, S. Liberman, J. Pinard, and A. Taleb, Phys. Rev.
Lett. 42,1404 (1979). 22. T. S. Luk, L. DiMauro, T. Bergeman, and H.
Metcalf, Phys.Rev. Lett. 47, 83 (1981). 23. D. A. Harmin, Phys.Rev. A 26,
2656 (1982). 24. W. E. Cooke and C. L. Cromer, Phys.Rev. A 32, 2725
(1985).

9 Magneticfields The effects of magneticfieldson Rydberg atoms have been

studied for more than 50 years. In fact, the experiments of Jenkins and Segre
on atomic diamagnetism were among thefirstin which the large size of the
Rydberg atoms was exploited,1 and when this topic was revisited by Garton
and Tomkins 30 years later they observed the unexpected quasi-Landau
resonances.2 Today an atom in a magneticfieldis oneofthe best systems
inwhichtosearchfor the quantum analogue of the onset of classical chaos.
Diamagnetism We canwrite the Hamiltonian for a H atom in a magneticfieldB
inthe z direction as H = �- - - + A(r) L �S+ � + S �B + - (x2 + y2)B2
(9.1) 2 Y 2 8 where B is given in units of 2.35 x 105 T, or 2.35 x 109 G,
A{r) is a function describing the spin orbit coupling, x, y, and zare the
Cartesian coordinates of the electron relative to the proton, and the electron-
proton distance Y = Jx2 + y2 + z2. Since(x2 + y2) <* n4,we canseethat the
ratioof the diamagnetic term to the linear magnetic field term is ~n4B, and for
afieldof 1 T n4B = 1at n =22. If n4B � 1 we may safely ignore the quadratic
diamagnetic term of Eq. (9.1). Inthiscasethe magneticeffects are of the
samesizeas they areinalow lying excited state of the same angular momentum,
with one smalldifference. Sincethe fine structure interval decreases as 1/n3,
due to the inverse r3 dependence of A(Y), the Paschen-Back regime in which
L and S are uncoupled is reached at lower B fields in Rydberg states than in
low lying states. If n4B >1 we cannot ignore the quadratic term of Eq. (9.1).
In this case it is reasonable to assume that L and Sare completely decoupled,
in which case S can often be completely ignored, as can the spin orbit
interaction A(Y)L,-S. The resulting Hamiltonian, H^-^ +^ +ltf + f)!}! (9.2)
has rotational symmetry about the z axis and preserves parity. Sincem and
parity are conserved, there are fewer coupled states than in an electric field,
in which 143

144 Rydberg atoms parity is not conserved, but the problem is non separable,
and, as a result, not amenable to analytic solution. If we rewrite the
diamagnetic term as HD = 1/8 r2 sin2 OB2,it is apparent that the non zero
matrix elements of Hu are those for which M = 0, �2, and Am = 0.3 There
is, however, no restriction on An.Explicit forms of the matrix elements are4
sin2 6\n'�m) = 2 ^ + \ ^ ^ <nt\?\n'Q (9.3a) and x <n�|r2|/i'(� + 2)). (9.3b)
The diagonal matrix element of r2 is the expectation value of (r2), given by5
1)]. (9.4) In H all � states of the same n are degenerate, and since the
diamagnetic interaction couples all the � states of the same parity by means
of the M = 2 matrix elements, � is not a good quantum number for any non-
zero field. If we consider an atom other than H, inwhich the low�states are
energetically removed from the high ( states, the first evidence of a
diamagnetic effect comes from the diagonal matrix element of //D, and
according to Eq. (9.2) we can expect a Rydberg level to be displaced from
its zero field energy by3 AW = � + � (ntmlr2 sin2 6\nZm). (9.5) 2 8 If we
approximate (r2�) by 5rc4/2, for the p Rydberg states, which are optically
accessible from the ground state, AW is given by3 A W = ^ +y ( l + mV- (9-
6) The m = �1 states are split by the linear shift and have twice the
diamagnetic shift of the m = 0 state. In an alkali atom the expressions of Eqs.
(9.5) and (9.6) for the shift are not valid for very high fields because the A�
= 2 matrix elements couple states of different �, leading to higher order
shifts. From second order perturbation theory it is clear that the shifts scale
as nnB4, indicating that the regime in which Eqs. (9.5) and (9.6) are likely to
bevalid is limited. Historically, the regime inwhich the diamagnetic effect
has produced eigenstates of mixed � is called the � mixing regime. In H it
starts at 5=0 and in an alkali, where it starts depends upon �. When the field
is further increased, so that the diamagnetic shifts exceed the n spacing the n
mixing regime is reached, in which the eigenstates have contributions from
different n as well as different �.13

Diamagnetism 145 The firstobservations of diamagnetism were made

byJenkins and Segre,1 who observed the absorption spectra of Na and K
using a 30 inch column of vapor located between the poles of a cyclotron
magnet. The magneticfieldwas 27 kG, enough to see clear diamagnetic effects
for n ~ 20. They only obtained good results with Na. With K they could not
observe the absorption of high series members unless the pressure was
sohigh that it broadened the lines. The Cooper minimum of K is at the
ionization limit, reducing the absorption of the high np states.6 In Na they
were able to observe the m =0 and �1 shifts and the m = �1 splittings given
by Eq. (9.6) up to n =25, at which point the shifts became larger than those
given by Eq. (9.6) due to �mixing. The Na np states lie just above the n-\
high � states and are forced up in energy by the A� = �2 diamagnetic
couplings. Although they were not able to resolve the other magnetic states,
for the higher n states they observed, n > 25, there were clearly visible tails
on the long wavelength sides of the lines indicative of other magnetic states
lying just below the np states. In experiments using laser techniques it has
been possible to resolve the magnetic levels and display the detailed
structure of the levels. An excellent example is the work of Zimmerman et
al.,7who excited Na Rydberg states in the fieldof a superconducting solenoid.
In their experiment a thermal beam of Na moves in the field direction and is
crossed by two pulsed laser beams. Thefirstis polarized perpendicular to the
field and is tuned to excite the atoms from the ground 3s1/2 state to the 3p3/2
ra; = 3/2 state, which contains only m = 1. The second laser, polarized
parallel to thefield,drivestransitions toeven parity m =1 Rydberg levels. The
Rydberg atoms are ionized by a 2kV/cm electricfieldpulse applied 1jus after
laser excitation, and the liberated electrons are accelerated to 10 keV and
detected with a surface barrier detector. Spectra are recorded at fixed
magnetic field strengths by scanning the wavelength of the second laser while
recording the electron signal. Their spectra, atfieldsof upto 6T are shown in
Fig. 9.1. Several features are apparent in the data of Fig. 9.1. First, the shifts
of the observed transition frequencies are quadratic in the magnetic field.
There is no linear shift since both the 3p state and m � \ Rydberg states have
the same linear shift. Second the agreement between the observed energy
level positions and those calculated by matrix diagonalization isexcellent.
Finally, the higher energy members of the set of diamagnetic states for a
given n have the largest oscillator strength and therefore the largest amount of
d character. Kleppner et al.4 have suggested that these two properties are
likely to be found together. The low � states have the largest classical outer
turning points and thus are likely to be prominent in the states with the largest
diamagnetic shifts. Fig. 9.1 also serves to bring out afinalpoint. The observed
m = 1 even parity levels of different ncross, atleast onthe
scaleofFig.9.1(b),whilethem = 0 evenparity levels,showninFig. 9.1(c), do
not. The m = 0levels contain the sstate with its large quantum defect. The
crossingofthe levels inthe nearly Hlikem = 1 evenparitystatessuggeststhat
there isa symmetry or conserved quality similar tothe Hlevelsinan electric
field.

146 Rydberg atoms 5 6 0 1 2 3 4 MAGNETIC FICLO <TESLA)

MAGNETIC FIELD (TESLA) Fig. 9.1 Diamagnetic structure of Na.
(a)Experimental excitation curves for even parity levels, m = 1,ras= i, in the
vicinity of n =28. A tunable laser was scanned across the energy range
displayed. The zero of energy is the ionization limit. Signals generated by
ionizing the excited atoms appear ashorizontal peaks. The horizontal scale
isquadratic in field. Calculated levels are overlaid in light lines. Some
discrepancies arepresent due to nonlinearity of the lasers, (b)Calculated
excitation curves, displayed linearly infield,(c) Same as (b), but foreven
parity m = 0 states. Note the large effect on anticrossings due to the presence
ofthe nondegenerate sstates (from ref. 7). As pointed out by Solov'ev,8 if the
magnetic field is low enough that the coulomb force is dominant, then there
exist approximate constants of the motion in addition to Lz and parity. The
first, A, is given by8"10 A = 4A2 - 5AZ2, (9.7) where A is the Runge-Lenz
vector A = p X L � r/r, which is directed along the semimajor axis of the
classical elliptical orbit. The second, Q, is given by8"10

Quasi Landau resonances 147 In a pure coulomb potential, i.e. in H or a high

m state of an alkali atom, the degeneracy of the orbits is so high that it is
relatively easy to form orbits which have approximately constant A and
Qfrom the many degenerate coulomb orbits.9 For A > 0the motion of
Aisprimarily arotation about the z axis,with Alying near the x,y plane. In
contrast, for A < 0the motion of A is a librational motion about the z axis
with A pointing near the z axis. The larger the value of A the more the motion
is concentrated in the x9y plane and the larger the diamagnetic energy shift.
For a specific n state the transition from A > 0 to A < 0 is marked by two
observable features. First, the spacing between levels, which is
approximately proportional to |A|, is a minimum for A ~ 0. Second, for Am =
0 transitions the oscillator strength from lower lyingstates is lowest to states
of A ~ 0and highest to the states with the highest and lowest values of A,
corresponding to Alying in the x,y plane and along the z axis, respectively.
These two points are shown clearly by calculations of Clark and Taylor11
and Cacciani etal.10 and by the experiment of Cacciani etal.10 They
observed single photon transitions from the ground state of Li to high lying Li
m = 0 odd parity states in a field of 1.94 T using a laser beam polarized
parallel to B. Their experimental approach was in essence the same as that of
Zimmerman et al. except that they drove one single photon transition from the
ground state rather than two. The largest quantum defect of the Li odd parity
ra = 0 states is that of the np states, which is 0.05, so these states are
reasonably hydrogenic. The experimental spectrum is shown in Fig. 9.2 along
with the calculated hydrogenic and Li spectra. Conventionally, the label K is
attached to the magnetic states, with K = 1being the highest energy state and
K = 15 being the lowest energy state in this case. For K = 1A = 3.5, for K =
15A = -1.0, and for K = 12 A = 0. In Fig. 9.2 it is evident that the lines are
weakest and closest together at K = 12, corresponding to A = 0. Quasi
Landau resonances Thirty years after the experiment of Jenkins and Segre,1
Garton and Tomkins2 observed the absorption spectrum of Ba vapor between
the poles of a 25 kG electromagnet. In zerofieldtheywere able to resolve
nplevels as high as75, and in the presence of the magnetic field they were
able to see individual diamagnetic levels as shown in Fig. 9.1. They were
also able to verify that forfieldsbelow the t mixing regime that the npm = 0
and �1 shifts were given by Eq. (9.6). The most interesting aspect of their
work, however, was that they were able to see quasiLandau resonances
above the ionization limit. An example isshown in Fig. 9.3,in which it is
clear that there is a modulation in the a spectrum, in which the light is
polarized perpendicular to the magnetic field, extending well above the zero
field limit. The energy spacing between the peaks is &W=-ha)c (9.9)

148 Rydberg atoms K=15 (a) -114.11 -110.34 cm"1 (b) ...1 0.26 cm"1 0
energy -114.112 -110.327 cm"1 (c) ...1 0.26cm" 0 energy -114.007 cm"1 Fig.
9.2 Diamagnetic structure of the n=31 multiplet in n excitation (B=1.94 T):
(a) experimental recording for Li; (b) calculated diamagnetic spectrum for
Li; (c) corresponding calculated spectrum for H (from ref. 10).

QuasiLandau resonances 149 n=40 n=40 Fig. 9.3 Diamagnetic Zeeman effect
in Ba: (a) zero-field, (b) o polarization with resonances extending into the
ionization continuum of (a) (from ref. 2). where a)c is the cyclotron
frequency eB/m. This spacing is 50% higher than the spacing of the Landau
levels of a free electron in a magnetic field and came as a surprise. While
direct diagonalization of the Hamiltonian matrix works well for
situationsinwhichthere isafinitenumber of states, as inFig. 9.1, itis clearly
hopeless to try it in this case. A useful WKB approach was proposed by
Edmonds12 and refined by Starace.13 Usingthe fact that azimuthal symmetry
exists, Starace writes the wavefunction of the spinless Rydberg electron in
cylindrical coordinates as13 /()e^. (9.10) p Using this wavefunction in the
Schroedinger equation leads to an equation for fiP,z), % ^ + % ^ + 2[W -
V(P,z)]f(P,z) = 0, (9.11) dp o z where

150 Rydberg atoms w -w-mB ~2~' and ' 2 p2 As pointed out by Edmonds
and Starace,12'13 the atoms are excited near the origin and can only escape
in the �z directions. The motion in the x,y plane is bound and is most likely
to be the source of the quasi Landau resonances. To find the locations of the
resonances it is adequate to ignore the z motion entirely and simply compute
the energy spectrum of the motion in x,y plane. Applying the Bohr-
Sommerfeld quantization condition leads to where n is an integer and m2 �
1/4 in the centrifugal term of Eq. (9.11) has been replaced by m2 to account
for the fact that the problem is two dimensional.14 In Eq. (9.12) pi and p2
are the inner and outer classical turning points. The spacing of the quasi-
Landau resonances is obtained by differentiating Eq. (9.12). Explicitly, dW
*;}�[ p2 p Evaluating Eq. (9.13) to obtain the resonance spacing, AW =
dW/dn, shows that at W = 0 AW = � = -B. (9.14) dn 2 As shown by Fig.
9.4, asthe energy is raised above the ionization limit the spacing between the
resonances slowly drops from (3/2)hwc to ha>c A most interesting question
is how to connect the quasi-Landau resonances of Fig. 9.3 to the diamagnetic
structure at lower energies, where it is understood in terms of individual
magnetic field states, as shown by Figs. 9.1 and 9.2. This question was
addressed byboth Fonck etal.15'16 and Economou etal.17 using laser
excitation of atomic vapor in a magnetic field and detection of the ions
produced by collisional ionization of the Rydberg atoms. Using two photon
excitation Fonck etal.15'16 observed clear m = 0 resonances in Sr similar to
the onesshownin Fig. 9.3 and less clear ones in Ba. Economou etal.17
observed the one photon Rb 5s �> np and two photon Rb 5s �> nd
transitions with circularly polarized light. In Fig. 9.5 we show the two
photon excitation spectrum of the Rb m = �2 states, the 5s�> nd series, in a
field of 55.5 kG. At n = 25 there is one readily identifiable resonance. As n
increases to 35 the single resonance splits into several peaks, which still fall
into recognizable groups. Finally at n = 42 the quasi Landau resonances at the
ionization limit are observed. Fig. 9.5 showsclearly that each of the quasi
Landau resonances ispart of a series of resonances each associated with

Quasi Landau resonances 151 10 kG 17 kG 25 kG -32 kG 40 Fig. 9.4 Plot of

the energy separation dWIdn against energy W above the threshold for six
values of the magneticfield,B = 10 kG, 17 kG, 25 kG, 32 kG, 40kG, 47 kG.
BothdW/dn and W are in units of the cyclotron energy, hcoc = h{eHlmc). As
W�>o�, (dW/dn)/hcoc^ 1 (from ref. 13). IONIZATION LIMIT (Ho�0)
5950 LASER WAVELENGTH A(A) Fig. 9.5 Data showing the m� �2,even
parity sublevels of Rb at 55.5kG (5s �> nd series at B = 0). Laser
wavelengths are plotted on the horizontal axis. The downward arrows
indicate the resonance positions predicted by a WKB model (from ref. 17).
asinglen state. Bymodifying theWKB expression ofEq. (9.12)togivethe
correct coulomb energies, Economou etal}1 were able to reproduce the
energies of the observed peaks from 350 cm"1below (corresponding to n =
25)to 20 cm"1 above the zerofieldionization limit. These experiments
showed clearly that n states evolve into the quasi-Landau resonances. In
higher resolution experiments, Gay et al.18 and Castro et al.19 showed that
itwas infact the highest energy diamagnetic stateswhichevolved into

152 Rydberg atoms 50"] I 1.92 I 1.84 I 1.83 I 1.77 I 1.74 I 1.63 I 155 I 1.54 I
1.52 I I I I I I I I I 60 40 20 0 BINDING ENERGY (cm"1) Fig. 9.6 Quasi-
Landau spectrum observed bylaser excitation andfield ionization of even
parity, m = �2 states of Na in a magnetic field of 4.2 T. The arrows indicate
the quasiLandau levels, the highest energy magnetic states ofeach principal
quantum number. The intermediate peaks are due to other levels. The
numbers between the arrowsgivethe level separation inunits of hcoc. A WKB
analysis predicts aspacing of 1.5 at W= 0, increasing with binding energy in
agreement with the data. Relative intensities are not reliable due to
laserfluctuations(from ref.19). the locations of the quasi-Landau resonances.
Gay et al.18 studied the nearly hydrogenic ra=3 states of Cs using a
thermionic diode in the magnetic field of a superconducting solenoid. Castro
et al.19 recorded the spectra of even parity Na m = � 2 states using the same
technique used to produce Fig. 9.1, except that both lasers were circularly
polarized in the same sense. In Fig. 9.6 their 4.2 T spectrum is shown. The
arrows show the highest lying magnetic energy levels of successive n,
starting with n = 35. These are clearly the dominant features of the spectrum,
and the interval between them, given in units of the cyclotron frequency,
clearly converges to 3/2, the quasi-Landau resonance interval, at the
zerofieldionization limit. In contrast to the m = � 2 spectrum shown in Fig.
9.6, in the analogous m = �1 spectra Castro et al.19 observed that many
magnetic levels of each n were excited approximately equally. Both of these
observations agree with the calculations of Clark and Taylor.11 If we now
return to the original experiment of Garton and Tomkins, we can ask why did
they not see resonances with Jtpolarization. More generally, when are quasi-
Landau resonances visible? The requirement is apparently similar to the
requirement for seeing strong field mixing resonances in an electric field.20
The energy ranges covered by the magnetic levels from each n are
overlapped at the ionization limit, and only if the oscillator strength is
concentrated in a few of the high energy magnetic states are the Landau
resonances visible. Exactly how the oscillator strength is distributed
variesfrom atom to atom and depends stronglyon the quantum defects of the
levels involved. However, for H like systems it ismore likely that the
requirement will be fulfilled for |Ara| = 1 than Am = 0single photon
transitions.

Quasi classical orbits 153 Quasi classical orbits While the two dimensional
treatment of Edmonds12 and Starace13 can be used to explain the
development of the quasi-Landau resonances from the bound statesit doesnot
giveanyinformation beyond thelocationsof the resonances. For example, it
does not tell us how wide the resonances are because motion in the z
direction is neglected. Furthermore, usingthisapproach we cannot tellwhether
or not there exist resonances corresponding to motion outside the x,y plane.
With these questions in mind Reinhardt21 suggested usingthe wavepacket
approach to the problem described in Chapter 8. The wave packet approach
successfully reproduces the quasi-Landau resonances,21 but more important,
provides away of looking for resonances which correspond to motions not
lying in the x,y plane. These wave packet notions proved to be invaluable in
understanding the later observations of Holle et al.,22 who observed the
quasi-Landau spectrum of H using aH beam propagating along amagnetic
field of up to 6 T. They excited H Is atoms towelldefined 2p m = 0 or 1
stateswith avacuum ultraviolet laser and then excitedfinalstates of m = 0,1, or
2 with asecond ultraviolet laser with alinewidth of 0.1-0.3cm"1 The laser
beams crossed the atomicbeam at aright angle. The H atomswere excited in
an electricfieldof 1 V/cmparallel to the magneticfield,and the electrons freed
bythe Rydberg atoms drifted out of theinteraction region and were
accelerated to 6keV to register on a surface barrier detector. When they
observed final states of m = 0, with both lasers polarized along the field
direction, they observed the familiar quasi-Landau resonances spaced at
3coc/2. In contrast, when they excited the m = 1finalstate via the intermediate
2p m = 1 state theseresonanceswere absent.
Onlyresonancesspacedby0.64a>cwere observed. In thefinalm = 0states the
oscillations at 3&>c/2 are evident, but in the final m = 1 states the
oscillations spaced by 0.64a>c are less so, although they are clearly evident
if the data are smoothed over 2 cm"1. Theperiodicity is even more apparent
in the Fourier transform of the spectrum, shown in Fig. 9.7. In Fig. 9.7(b) the
peak corresponding to the resonance spacing of 0.64a)c is readily apparent.
The origin of this resonance was identified by extending Reinhardt's21 wave
packet notion. Realizing that the wave packet evolves along the classical
trajectories of the electron, Holle etal.searched for classical trajectories in
which the electron leaving the origin returned in a time of 9.5 ps, l/2jr
(0.64&>c). The orbit they discovered which has this return, or recurrence,
time does not lie in the x,y plane, and thus cannot be predicted by the
Edmonds-Starace approach.1213 It is clear that it is only possible to observe
these resonances if their recurrence times are shorter than the coherence time
of the exciting laser. With this point in mind, Main et al.23 made afivefold
improvement in their spectral resolution and were able to see new
resonances in the Fourier transform spectrum with longer recurrence times,
as shown in Fig. 9.8. Tc is the recurrence time for a cyclotron orbit. The
spectrum vs tuning energy is not shown since it is composed of

154 Rydberg atoms 10 20 30 40 TIME [1O'12 sec] Fig. 9.7 Fourier

transforms of two photon, resonant, excitation spectra of the H atom Baimer
series around the ionization limit in a magneticfieldof strength B =6T, excited
through individually selected magnetic substates m = 0andm = +1 of the n
=2state to final m states of even parity: (a) m =0, (b) m � +1 (c) m = +2,
plus some admixture (-25%) of m =0. The resolution is ^0.3 cm"1 (from ref.
22). apparently random lines. Nonetheless, the Fourier transform spectrum of
Fig. 9.8 shows clear peakswhich may be correlated with
classicalorbitswhichreturn to the origin. In Fig. 9.9 we showthe orbits which
correspond to the peaks in the Fourier spectrum of Fig. 9.8. As shown by Fig.
9.9 the original quasi-Landau resonances, v = 1, are the only ones which
have orbits in the x,y plane.

Quasi classical orbits 155 Time (Tc) Fig. 9.8 Fourier transforms of H
spectra obtained in a magnetic field of 5.96 T with resolution 0.07 cm"1: (a)
initial state 2p m = 0, final state m = 0 even parity states; (b) initial state 2p
m = -1 final state m = -1 even parity states. The squared value of the absolute
value is plotted in both cases. The circled numbers correspond to the
classical orbits depicted in Fig. 9.9 (from ref. 23). An interesting hybrid way
of calculating the spectrum is one used by Du and Delos.24They start with
aquantum mechanical wave packet at the origin and letit propagate to r =
50a0, where they use the normals of phase fronts of the wave packets to
define classical electron trajectories. These classical trajectories are then
followed. Some of the trajectories are reflected back to the origin, and when

zxl0 3a o. y*10 3a o p x1 0J ao - 2 - 1 1 2 p x ] 0 3 a -o. 5 15 x *10 3 n Fig

.9. 9 (a),(b )Calculate d close d trajectorie so felectro n motio n at energ y W
= 0, correspondin g to the first seve n resonance sshow ni nFig . 9.8 .Th
efinal stat ei si nal lcase sm = 0. Projection sont oth ep,z plan ear edepicte
dan dth eequipotentia lW = 0i sshown ,(c )Close d trajector y of typ e3 at W
= 0show ni nprojectio n ont oth ex,y plane .Th efinal stat ei sagai nm = 0(fro
m ref .23) .

Quasi classical orbits 157 they come back to r = 50a0 they are used to
construct incoming wavepacket phase fronts. The resulting incoming wave
packet can interfere constructively or destructively with the outgoing wave
packet at the origin, leading to total wavefunctions of varying amplitude, and
photoionization cross sections which vary accordingly. Their results are in
excellent agreement with the experimental results of Main et al.23 The fact
that the resonances in the Fourier transform spectrum of Fig. 9.8 can be
predicted byfinding classical orbits which return to the origin suggests a
better way of looking for the classical orbits. Wintgen25 and Wintgen and
Friedrich26 suggested that if the classical Hamltonian is given by H -�-7 +
i�v- (915 � then using the scaled variables f = y2/3r (9.16a) and P =
y"1/3P, (9.16b) where y = B/2.35 x 105T (y is B in atomic units), //(f,p,y) =
y2/3//(r,p,y=l). (9.17) Applying the semiclassical quantization condition to
the action around a closed classical orbit yields25"27 �lpp dp + pz dz =
ny1/3 = Q, (9.18) where Q is the scaled action of the closed orbit. Since the
action scales with the scaled energy, as W � Wy~2/3, if the scaled energy is
fixed, the spectrum of possible actions is simply an integer times y113. Holle
etal.27observed the scaled energy spectrum bykeeping W = Wy~2/3 fixed
while scanning B~1/3 linearly, or equivalently y~113 linearly. Their
experimental spectrum, when Fourier transformed, leads to the Fourier
transform spectrum of Fig. 9.10, in which clear peaks separated by y~1/3 are
visible. This spectrum is at a scaled energy W = -0.45, corresponding to an
energy range -77.7 cm"1 < W < -54.3 cm"1 and 5.19 T > B > 3.03 T. At this
scaled energy the classical orbits are regular and the spectrum is well
behaved. At higher scaled energies, W 0.1, the peaks of Fig. 9.10 break into
several peaks and the regularity of Fig. 9.10 disappears, which is often
termed the onset of classical chaos. While the success of scaled energy
spectroscopy suggests that the behavior of atoms does become classically
chaotic near the ionization limit, higher resolution reveals a surprisingly
orderly structure. Iu et al2S have studied the odd parity Li m = 0 states in a
beam travelling in the direction of the magnetic field. They

158 Rydberg atoms -82.93cm"1 < W > -56.77cm"1 5.72 T < B > 321,1 (b)
Fig. 9.10 (a) Scaled energy spectrum at W= -0.45 as a function of y 1/3 .
Range of excitation energy -77.7 cm"1 < W < -54.3 cm"1 and field strength
5.19 T > B >3.03 T. (b) Fourier transformed action spectrum of (a); closed
orbits correlated to respective resonances inp,z projection; z coordinate
vertical (from ref. 27). excited the Li atoms from the ground 2s state to the 3s
state using two photon excitation, and then drove the single photon 3s-^np
transitions with 30 MHz resolution. They observed surprisingly regular
structure in the vicinity of the ionization limit, a regime which is classically
chaotic according to the work of Holle et al.21 In Fig. 9.11, we show their
spectra, taken from 1 cm"1 below the ionization limit to 3cm"1 above the
limit. As shown by Fig. 9.11, the spectra consist of strong lines, which
increase rapidly in energy with field, and weaker lines which increase less
rapidly with field. The simple fact that the spectrum isso apparently regular
is remarkable. To explain the spectrum Iu et al. proposed that the adiabatic
model of Friedrich29 could be used. The essential idea is that the

Quasi classical orbits 159 5.95 6.00 Reid [T] 6.15 6.09 6.11 Reid [T] 6.17
Fig.9.11 Energy levelmapofoddparitym = 0 statesofLiin
anappliedmagneticfield.The mapsare created byscanning alaser
atsuccessively higherfields:Horizontalpeaksare field ionization signals. (The
gaps are regions which were skipped during data collection.) The structure is
most easily seen by viewing close to the plane from the left. Reduced-
termvalue plots for the outlined regions are shown at the right. (a),(b) np = 1;
(c) np =0. With these coordinates a progression of Rydberg states appears as
a series of lines with unit separation. The ionization threshold in
thefieldoccurs at approximately +2.8 cm"1 (from ref. 28).
160 Rydberg atoms electron motion is separated into the rapid motion in the
x,y plane and a slow motion along the z axis. With this separation the
energies are given by28 W(np,vz) = (np + 1/2)B - ^ , (9.19) where np is the
quantum number of the Landau level and vz is the effective quantum number
of the motion in the one dimensional coulomb potential in the z direction.30
Although Friedrich developed the model for higher magneticfields,it seems
to match the observed spectrum remarkably well in some cases. In the
sections on the right side of Fig. 9.11 the energy of the Landau level,
corresponding to np = 0 or 1 has been removed, leaving only the energy of
the coulomb motion in the z direction. The fact that the levels match the
coulomb spacing is apparent, although the quantum defect seems to decrease
substantially with magnetic field. While this theoretical description is
qualitative, it offers the promise that the spectrum in high magnetic fields can
be explained quantum mechanically. References 1. F. A. Jenkins and E.
Segre, Phys.Rev. 5552 (1939). 2. W. R. S. Garton and F. S. Tomkins,
Astrophys. J. 158, 839 (1969). 3. L. I.Schiff and H. Snyder, Phys.Rev. A 55,
59 (1939). 4. D. Kleppner, M. G. Littman, and M. L. Zimmerman, in Rydberg
States ofAtoms of Molecules, eds. R.F. Stebbings and F.B. Dunning
(Cambridge Univ. Press, Cambridge, 1983.) 5. H. A. Bethe and E. A.
Salpeter, Quantum Mechanics of One and Two Electron Atoms (Plenum,
New York, 1977.) 6. W. Sandner, T. F. Gallagher, K. A. Safinya, and F.
Gounand, Phys.Rev. A 23, 2732 (1981). 7. M. L. Zimmerman, J. C. Castro,
and D. Kleppner, Phys.Rev. Lett. 40,1083 (1978). 8. E. A. Solov'ev, JETP
Lett.34, 265(1981). 9. J.C. Gay and D. Delande, Comm At. Mol. Phys.
13,275(1983). 10. P. Cacciani, E. Luc-Koenig, J. Pinard, C. Thomas, and S.
Liberman, Phys.Rev. Lett. 56, 1124(1986). 11. C. W. Clark and K. T. Taylor,
/. Phys B 15,1175(1982). 12. A. R. Edmonds, /. Phys. (Paris) 31, C4 (1970).
13. A. F. Starace, /. Phys. B 6, 585 (1973). 14. O. Akimoto and H.
Hasegawa, /. Phys.Soc. Jpn. 22,181 (1967). 15. R. J. Fonck, D. H. Tracy, D.
C. Wright, and F. S. Tomkins, Phys.Rev. Lett. 40,1366 (1978). 16. R. J.
Fonck, F. L. Roesler, D. H. Tracy, and F. S. Tomkins, Phys.Rev. A 21, 861
(1980). 17. N. P. Economou, R. R. Freeman, and P. F. Liao, Phys.Rev. A 18,
2506 (1978). 18. J.C. Gay, D. Delande, and F. Biraben, /. Phys. B 13,
L729(1980). 19. J.C. Castro, M. L. Zimmerman, R. G. Hulet, D. Kleppner,
and R. R. Freeman, Phys. Rev. Le�. 45,1780(1980). 20. E. Luc-Koenig and
A. Bachelier, Phys.Rev. Lett.43,921 (1979). 21. W. P. Reinhardt, /. Phys.
B16, 635 (1983). 22. A. Holle, G. Wiebusch, J. Main, B. Hager, H. Rottke,
and K. H. Welge, Phys.Rev. Lett. 56, 2594(1986). 23. J.Main, G. Wiebusch,
A. Holle and K. H. Welge, Phys.Rev. Lett.57,2789(1986).

References 161 24. M. L. Du and J. B. Delos, Phys.Rev. Lett. 58, 1731

(1987). 25. D. Wintgen, Phys. Rev. Lett. 58, 1589 (1987). 26. D. Wintgen and
H. Friedrich, Phys. Rev. A 36, 131 (1987). 27. A. Holle, J. Main, G.
Wiebusch, H. Rottke, and K. H. Welge, Phys. Rev. Lett. 61, 161 (1988). 28.
C. Iu, G. R. Welch, M. M. Kash, L. Hsu, and D. Kleppner, Phys.Rev. Lett. 63,
1133(1989). 29. H. Friedrich, Phys. Rev. A 26, 1827(1982). 30. R. Loudon,
Am. J. Phys. 27, 649 (1959).

10 Microwave excitation and ionization In our encounter with ionization

bystatic and quasi-static fields we saw that there are substantial differences
between the ionization of H or H like atoms and nonhydrogenic atoms. These
differences arise from the fact that in an atom other than H the presence of the
ionic core couples states which are not coupled in H. Thiscoupling results
inthe avoided crossingsof bound Stark energy levels and the coupling of
nominally stable Stark levels to the continuum of red shifted Stark levels of
highern states. Differences existin both staticandpulsedfield ionization, and it
is hardly a surprise that microwave ionization of hydrogenic atoms differs
from that of nonhydrogenic atoms. For a> < l/n3, where w is the microwave
frequency, hydrogenic atoms of \m\ �n ionize at afield of l/9n 4 , asthe red
Stark states do in a static field.1 Low m states of alkali atoms, however,
ionize at the radically different microwave field given by2'3 E = l/3n5.
(10.1) We beginbybriefly summarizing thetwo experimental
techniquesusedtostudy microwave excitation and ionization. Thefirst
technique is theuseoffast beams of H and He Rydberg atoms. An apparatus
used to study the microwave ionization of H andHe is showninFig.
10.1.3Adistribution of Rydberg statesisproducedby charge exchange of a
continuous beam of H+ or He+ ions, and high lying states, aboven � 10,are
removed byfield ionization toproducewhatis labelled the atom beam inFig.
10.1. Selected single Rydberg states arethen populated byexcitation of atoms
in typically an n = 7 state, first to an n = 10 state and then to a higher ATOM
BEAM MIRROR Fig. 10.1 Schematic drawing of afast beam apparatus. A
fast atomic beam enters from the left andis excited sequentially by two
different CO2lasersinelectricfieldregionsF1 andF3, respectively. F2 avoids a
zerofieldregion between them. Ions produced by highly excited atoms being
ionized in the biased microwave cavity are energy selected and detected bya
Johnston particle multiplier (not shown). The output signal is detected in
phase with the mechanically chopped F1 laser beam (from ref. 3). 162

Microwave excitation and ionization 163 Signal Out Fig. 10.2 Major
components of athermal atomicbeam apparatus for microwave ionization
experiments,the atomic source, the microwave cavity, and the electron
multiplier. The microwave cavity is shown sliced in half. The Cu septum
bisects the height of the cavity. Two holes of diameter 1.3 mm are drilled in
the side walls to admit the collinear laser and Na atomicbeams, and a 1
mmhole inthe top ofthe cavity allowsNa+ resultingfrom a field ionization of
Na to be extracted. Note the slots for pumping (from ref. 4). Rydberg state
using two CO2 lasers. The Rydberg atoms pass through a microwave cavity,
where they are exposed to a strong microwave field. Due to the high
velocities of the energetic, �10keV, H and He beams, the atoms pass through
the microwave field region in a short time. For example a 10 keV H beam
has a velocity of 1.3 x 108 cm/s and passes through a 5 cm long cavity in 40
ns. Downstream from the cavity the beam is analyzed, usually by applying an
electric field, to see if ionization or transitions to other states have occurred.
In these experiments the excitation, interaction with the microwave field, and
final state analysis are separated spatially, as shown in Fig. 10.1. The second
approach is to use thermal beams of alkali atoms as shown in Fig. 10.2.4 A
beam of alkali atoms passes into a microwave cavity where the atoms are
excited by pulsed dye lasers to a Rydberg state. A1 JUS pulse of microwave
power is then injected into the cavity. After the microwave pulse a high
voltage pulse is applied to the septum, or plate, inside the cavity to analyze
the final states after interaction with the microwaves. By adjusting the
voltage pulse it is possible to detect separately atoms which have and have
not been ionized or to analyze by selective field ionization the final states of
atoms which have made transitions to other bound states. As we shall see,
microwave ionization can be thought of as a multiphoton absorption or as a
process driven by a time varying field. We first discuss microwave
ionization of alkali atoms, which can be described by the notions used to
describe pulsed field ionization. To show the connection between the time
varying field point of view and the photon absorption point of view we then
discuss
164 Rydberg atoms n 18 20 22 24 26 28 303234363840 1000 Fig. 10.315
GHzmicrowaveionizationfieldsforthe \m\< 1 components oftheNa (n + l)s (O)
and nd (�) states. Ionization fields of the \m\ =2 components of the nd states
(�) (fromref. 4). microwave multiphoton resonance experiments, which
allow us to refine our understanding of microwave ionization. We then
discuss ionization of H like atoms. Finally, ionization by circularly polarized
fields is described. Microwave ionization of nonhydrogenic atoms As noted
earlier, nonhydrogenic atoms ionize at amicrowave field of E ~ l/3n5, a field
substantially less than the static field required for ionization, E = l/16n4.
This point is made quite graphically in Fig. 10.3, a plot of the observed
fields for 50% ionization of the Na nd \m\ =0 and 1 states by a 15 GHz
microwave field pulse lasting 500 ns.2 Also shown in Fig. 10.3 are the very
different ionization fields of the approximately hydrogenic Na \m\ = 2 states.
Although these fields are often called microwave ionization thresholds they
are not nearly as sharp as the pulsed field thresholds shown in Fig. 7.4, a
point which ismade by Fig. 10.4, a plot of the microwave ionization
threshold of the Na 20s state.4 The complementary signals of Fig. 10.4 are
obtained by detecting the atoms which have and have not been ionized by the
microwaves. In addition to Na, experiments have been done in other atoms,
He, Li, and Ba,1'5'6 and these experiments also show that ionization
generally occurs when E ~ l/3n5.

Microwave ionization of nonhydrogenie atoms 165 MICROWAVE FIELD

AMPLITUDE (V/cm) 300 400 500 600 700 8 4 0 MICROWAVE
ATTENUATION (db) Fig. 10.4 Fieldionization signal (�) and15
GHzmicrowave ionization signal(�) forthe Na 20sstate showing
theionization threshold asboth thedisappearance ofthefieldionization signal
and the appearance of the microwave ionization signal with increasing
microwave power (decreasing attenuation). For convenience the microwave-
field amplitude is also given (from ref. 4). The fact that atoms can be ionized
by a field so far below the classical field required for ionization, as shown
by Fig. 10.3, suggests that the microwave field induces transitions to higher
lying states which can be directly field ionized by the microwave field, and
that what we observe as the microwave ionization threshold field is in fact
the field required to drive the rate limiting step. The fact that the threshold
field is so close to E = l/3n5, the field of the avoided level crossing between
the extreme n and n + 1 Stark levels, suggests that the rate limiting step is a
Landau-Zener transition between these two levels at their avoided crossing.
It isstraightforward to generalize the notions previously used to describe
pulsed field ionization to understand microwave ionization as a process
driven by a temporally varying field. As an example we consider the
ionization of an atom initially excited to the Na 20d state with \m\ = 0. In Fig.
10.5 we show the energy levels of the Rydberg states in positive and
negative electric fields.2 For clarity, we have not shown the s and p states,
with quantum defects 1.35 and 0.85, nor have we shown all the Stark energy
levels in detail, but only shaded the field energy region in which they lie.
Consider applying a pulse of microwaves to atoms initially excited to the Na
20d state. In the presence of the field the good quantum states are the Stark
states, and the microwave field rapidly induces transitions

166 Rydberg atoms Classical Ionization Limit -2000 -1000 0 1000 2000 Fig.
10.5Energy level diagram showing the mechanism bywhich ann = 20atom is
ionized by amicrowave field of 700 V/cm amplitude. An atom initially
excited to the20d state is brought to the point at which the n = 20 and 21
Stark manifolds intersect. Atthispointthe atom makesa Landau-Zener
transition, as shownbytheinset,tothen =21 Stark manifold and on subsequent
microwave cycles makes further upward transitions as shown by the bold
arrow. The process terminates when the atom reaches a sufficiently high n
state that the microwave field itself ionizes the atom. This occurs atpoint B in
this example (from ref. 4). between the n = 20 \m\ =0 Stark states at the
avoided crossings at zerofield.Thus roughly 5% of the atoms are in each
Stark state at any given time. If the microwavefield reaches E = l/3n5, atoms
in the highest energy n = 20Stark state are brought to the avoided crossing
with the lowest lying n =21 Stark state at E = l/3n5, and these atoms can
make a Landau-Zener transition to the lowest n = 21 Stark state at the
avoided crossing asshown in theinset of Fig. 10.5. We shall shortly return to
consider the Landau-Zener transition in more detail, but for a moment we
assume that the Landau-Zener transition does occur. It is evident that
afieldadequate to drive the n�> n + 1 transition is adequate todrive the n + 1
�> n + 2transition anda sequence of similar transitions to higher lying states,
culminating in ionization when it is classically allowed, as shown byFig.
10.5. On the other hand the field is not strong enough to drive the transition to
n = 19, so atoms donotmake transitions to lower n. In sum themicrowave
field drives transitions to higher n, resulting ultimately inionization, and the
ionization field is determined by the rate limiting step, then�� n +
1transition.

Microwave ionization of nonhydrogenie atoms 167 LU Field Fig. 10.6

Possible outcomes of a double traversal of an avoided crossing starting from
A. Veryslow changeinfield A�> B -� A. Veryfast chargeinfield A�> C�>
A. Intermediate rate of change A�> Band C �> D and A. The Landau-Zener
transition can be understood by considering the example shown in Fig. 10.6
in which, at its peak, the microwave field goes slightly past the avoided
crossing. If the field traverses the crossing too slowly an atom initially in the
lower state A remains at all times on the lower energy level reaching point B
at the peak of the fieldthen returning to A when thefieldreturns to its original
value. If thefield traverses the crossing too rapidly an atom makes adiabatic
transition to point C on the upper level on the increase of the field and returns
to A when the fieldreturns to its initial value. On the other hand, if the
crossing is traversed in a time comparable to the inverse of its magnitude,
when the field increases to its peak there is a probability of the atom's being
in both the upper and lower states, at points B and C, and after the field
returns through the crossing there is a probability of the atom's having made
the transition to point D. To achieve a high, 50%, transition probability the
field must reach the avoided crossing, but the transitions still occur, with
lower probability, even if the field does not reach the crossing. Since there
are many field cycles in the microwave pulse the transition probability need
not be high on a single cycle, and the field does not actually need to reach the
crossing field, E = l/3n5. Pillet etal. ,4using the numerical method of
Rubbmark et al.,7 have calculated transition probabilities between extreme
Stark states due to a single half cycle of the microwave field. At 15 GHz
these calculations indicate that the n �> n + 1 transition probability is �1%
when E = 85% of the crossing field, i.e. E = 1/3.5n5, in reasonable
agreement with the experimental observations shown in Fig. 10.3. These
calculations also show that, if the field amplitude is fixed at l/3n5, the
maximum transition probability occurs when the microwave frequency co is
comparable to co0, the size of the avoided crossing between the n and n 4- 1
levels. However, the transition probability is negligible if co � co0 or co �
co0. The requirement co ~ co0 for the Landau-Zener transition is consistent
with the fact that ionization at E = l/3n5 does not occur in H like states for
which the avoided crossings are vanishingly small, i.e. co � co0.

168 Rydberg atoms While the above Landau-Zener description gives an

appealing picture of ionization by means of the avoided crossings, it is an
oversimplification of the actual process in that it is based on a single
microwave cycle and a single pair of levels. Considering only a single cycle
ignores the possibility of coherence between field cycles and the
contributions of nonextreme Stark states with avoided crossings at E > 5
Microwave multiphoton transitions The field at which ionization occurs in
nonhydrogenic systems when co � 1/n 3 is determined bythe rate limiting n
�> n + 1 transition, whichwe have described as a Landau-Zener transition.
Since this transition is between real states it should, in principle, be possible
to observe it as resonant photon absorption. However, in microwave
ionization it is difficult to separate the rate limiting step from the ensuing
ionization. Instead, it isuseful to consider transitions between an isolated pair
of levels. An example is provided by the K atom. As shown by Fig. 10.7 for
n = 16, the K (n + 2)s states intersect the n Stark manifolds at the field Ec ~
l/10tt5, far below thefieldrequired for microwave ionization.8'9 Thus it is in
principle possible to examine only the transition from the (n + 2)s states to
the (n,k) Stark states without having the succession of transitions to higher n
aswell. For simplicity we shall use the convention of labelling each (n,k)
Stark state by the zero field n( state to which it is adiabatically connected.
Thus the (16,3) state is adiabatically connected to the 16f state and is the
lowest member of the n = 16 Stark manifold for E < 1/lOn5. There are two
additional attractive features of these K transitions. First, the locations and
widths of the avoided crossings between the (n + 2)s and (n,k) states have
been measured by anticrossing spectroscopy.9 Second, the static Stark shifts
of both levels are nearly linear, so the avoided crossing of these two levels
is a good approximation to the n �� n + 1 avoided level crossing at E =
l/3n5. The experiment is done using the apparatus shown in Fig. 10.2. Two
tunable dye lasers are used to excite K atoms in a beam to the (n + 2)s state.
A static field can be applied to vary the separation between the 18s and
(16,/:) states. The atoms are exposed to a microwave pulse, and subsequently
to a field ionization pulse which ionizes atoms in the (n,k) Stark state, but not
those in the (n + 2)s state. The (n,k) field ionization signal is then monitored
as either the microwave field amplitude or the static field is swept over
many shots of the laser. To the extent that both Stark shifts are linear, the
static field alters the energy separation between the two states, but not their
wavefunctions. The similarity of the K (n + 2)s�� (n,k) transitions to the
n�> n + 1 transitions of microwave ionization isverified by measuring the
number of atoms making the former transition as a function of the microwave
field amplitude.10 As the

Microwave multiphoton transitions 169 -410 -420 E MERG Y 1 UJ -430

-440 -450 200 400 600 800 1000 1200 ELECTRIC FIELD (V/cm) Fig. 10.7
Relevant energy levels of K near the n = 16 Stark manifold. The Stark
manifold levels are labeled (n,k), where k is the value of � towhich the
stark state adiabatically connects at zero field. Only thelowest two
andhighest energy manifold states areshown. The laser excitation to the
18sstate isshown bythelong vertical arrow. The 18s �> (16,3) multiphoton
rf transitions arerepresented bythebold arrows. Note that these transitions are
evenly spaced in static field, and that transitions requiring more photons
occur at progressively lower static fields. For clarity, the rf photon energy
shown in the figure is approximately 5times its actual energy (from ref. 8).
microwave field amplitude is increased, a sharp increase in the number of
atoms making the (n + 2)s^ (n,k) transition is observed when the microwave
field amplitude reaches a threshold field near the field, Ec, of the avoided
crossing between the (n + 2)s and lowest (n,k) Stark state. For example,
when K 18satoms are exposed to a 1juspulse of 9.3 GHz microwaves of
variable amplitude, the threshold is at 775 V/cm, close to the avoided
crossing field, Ec = 753 V/cm. For 15 < n < 20, the observed microwave
threshold fields vary from 1.0Ec to 1.6EC. Although the variation from 1.0Ec
to 1.6EC is hard to reconcile with the Landau-Zener theory, a threshold field
near Ec is expected. There is one exception, the K 19s �> (17,/r) transition,
in which a non-monotonic increase in the transition probability is observed
with 9.3 GHz field amplitude, beginning at fields well below the avoided
crossing field, Ec = 546 V/cm, asshown byFig. 10.8. Similar observations
were made inHe byvan de Water etal.n This observation is, of course,
incompatible with the Landau-Zener picture but can be explained with a
multiphoton description.

170 Rydberg atoms -z. o 405 510 575 645 725 810 910 1020 MICROWAVE
FIELD (V/cm) Fig. 10.8 The K 19s-> (17,k) transition as afunction of
microwave field. The 19s state is excited by the lasers, a microwave field is
applied, and atoms which have made the transition to the (17,A:) states are
detected by field ionization. The static field avoided crossing is at 546V/cm
(from ref. 10). Sweeping the static field, which alters the energy spacing
between the levels, while keeping the microwave field fixed leads to the
observation of resonant multiphoton transitions. An example, shown in Fig.
10.9, is the set of the 18s �>(16,fc) transitions observed by sweeping the
static field with different strengths of the 10.35 GHz microwave field.8 The
sequence of 18s �> (16,3) transitions 25 V/cm apart is quite evident. At the
top of Fig. 9 there is a scale in terms of the number of 10.35 GHz photons
required to drive the 18s�� (16,3) transition. At low microwave fields
transitions involving only a few photons are observed, at high
staticfields,near thefieldof the avoided crossing. Asthe microwavefieldis
increased more transitions are observed, at progressively lower static fields,
until at the highest microwave fields the sequence of resonant transitions
extends to zero static field. As shown by the scale at the top of Fig. 10.9 the
18s �> (16,3) transition nearest zero static field corresponds to the
absorption of 28 photons. Careful inspection of Fig. 10.9 reveals three
interesting features. First, most of the 18s �� (16,3) resonances
corresponding to the absorption of agiven number of

Microwave multiphoton transitions 111 PHOTONS 14 13 12 11 10 9 8 7 6 5

4 3 2 1 (3 CO (a) 3.4 V/cm 10.2 V/cm 114 V/cm 350 400 450 500 550 600
650 700 750 STATIC FIELD (V/cm) PHOTONS 28 27 26 25 24 23 22 21 20
19 18 17 16 15 (b) 0 50 100 150 200 250 300 350 STATIC FIELD (V/cm)
Fig. 10.9 (a) K 18s�> (16,3) one tofourteen photon transitions observed as
the staticfieldis scanned from 350-750 V/cmfor the 10.353 GHz microwave
fields indicated above each trace (3.4�190 V/cm). The regularity of the
progression is quite apparent. Note the extra resonances in the 142 V/cm
microwave field trace. These are due to 18s �> (16,4) transitions, (b) 18s
�> (16,3) 15-28 photon transitions observed as the static field is scanned
from 0 to 350 V/cm. Note thecongestion of the 410-V/cm trace at static fields
above �200 V/cm, dueto many overlapping 18s �> (16,/:) transitions (from
ref. 8).
172 Rydberg atoms W1 > o LJJ LU W2 |2> j I N 0 Es -Em w Es Ec
ELECTRIC FIELD Fig. 10.10 Multiphoton resonant transition at anavoided
crossingfrom thephoton and field points of view. The solid curves are the
avoiding levels and the dashed lines are the levels which cross when the
coupling is ignored. The static field Es gives rise to a six photon resonant
transition, indicated by the stacked arrows. The range of the electric field
variation is shownfor thecase inwhichthepeakfieldE$+ �^w
exactlyreachesthecrossing is exaggerated for clarity (from ref. 10).
microwave photons are not shifted in static field as the microwave field is
increased. Only the resonances at very low static fields, requiring the highest
microwave powers, shift as the microwave field is increased, moving to
higher static field. Second, in the spectra at microwave fields of 114 V/cm
and higher there are resonances which do not match the 25 V/cm interval of
the 18s �> (16,3) sequence. The "extra" resonances at microwave fields of
114-190V/cm appear at intervals of 34 V/cm and are the 18s-^ (16,4)
transitions. Especially for high microwave fields, as the static field is
increased, resonances to (16,/:) states of k > 3 are observed, and these
overlapping resonances form an effective threshold field for the 18s�>
(16,fc) transitions.Third, beyondthetwophoton transition, the number of 18s-
� (16,3) transitions observed increases linearly with the microwave field.
The data of Fig. 10.9 show clearly that the K (n + 2)s�>(n,k) transitions are
multiphoton transitions. On the other hand, most of the data obtained by
sweeping the microwave field agree qualitatively with the Landau-Zener
description. Toreconcilethese apparently different descriptionswe
considertheproblem shown inFig. 10.10,in which there are two states 1 and
2, with alinear Stark shift and no Stark shift respectively.10 States 1 and 2
are coupled by V, the core

Microwave multiphoton transitions 173 interaction, which is time

independent. Without Vlevels 1 and 2would cross at the field Ec, but due to
the coupling V there is an avoided crossing of size a)0 at the static field Ec.
First, we shall approach the problem as atransition driven by the time
variation of a quasi-static field, in other words, in the Landau-Zener terms
we used to describe microwave ionization. We can write the Schroedinger
equation for this problem as 0 dt (10.2) As we have described the problem
above, the full Hamiltonian is given by H=HH+V+HEs + HEmw (10.3)
where HH isthe Hamiltonian for the H atom in zero field, Vis the core
interaction and HEs and HEmw represent the interactions with the static and
microwavefields. If we ignore for a moment the core interaction, V, and
imagine that the field is static, i.e. use as the Hamiltonian in Eq. (10.2), H =
HH + HEs, there are two solutions to Eq. (10.2) corresponding to the two
eigenstates 1and 2; ^(r, t) = Vi(r)e-i(wi-^s)' (10.4a) V2(r,0 = V2(r)e-iw2',
(10.4b) where W1 and W2 are the zero field energies of states 1 and 2 and k
is the permanent dipole moment of state 1. Following common usage we have
used k as both a state label and as a permanent dipole moment. The total
wavefunction is given by V(r,f) = *Vi(r,0 + Hi(?J) (10.5) where a and b are
time independent and a2 + b2 = 1. In this approximation the energy levels
cross at Ec = {W1 � W2)/k. If we add the coupling V to the Hamiltonian,
leading to H = HH + HEs + V, the coupling lifts the degeneracy at Ec, and the
two levels are separated in energy by <y0, defined by ^ = <2|V|l> = /
Vl(r)VVi(r) dr. (10.6) Now imagine an experiment with states 1 and 2
analogous to the Kexperiments. Initially the atoms are in state 2 in a static
field, the microwave pulse is applied, and afterwards the atoms can be in
state 1 or state 2. During the microwave pulse we must use the full
Hamiltonian of Eqs. (10.4). We can no longer represent the wavefunction by
Eq. (10.5), but rather by V(r,0 = 7\(0Vi(r) + T2(t) </<2(r). (10.7) Using the
wavefunction of Eq. (10.7) in the Schroedinger equation, Eq. (10.2), leads to
two coupled equations for Tx{i) and T2(f). Explicitly i fx = [Wt - kEWTi +
bT2 (10.8a)

174 Rydberg atoms Field Fig. 10.11 Schematic picture of a multicycle

Landau-Zener transition. In combined static and microwave fields the
oscillating field brings the atom to the avoided crossing on successive
cycles, and the transition amplitudes due to successive cycles add, leading to
interference, or resonances (from ref. 12). i f2 = bTx + W2T2. (10.8b) The
total electric field is given by E(t) = Es + Emw cos ot, (10.9) where Es, and
�mw are the static and microwave field amplitudes. To obtain a solution to
Eqs. (10.8), they can be integrated numerically, using the method of
Rubbmark et al.1 Letting the time evolve from cot = - nil to nil yields results
analogous to the half cycle results of Pillet et al.4 When the time interval is
extended to more than three cycles, resonances appear in the transition
probabilitywhen the staticfieldhasvaluesfor which the energy difference
between the two states in the static field alone is an integral multiple of the
microwave frequency.10 After many cycles the calculated lineshapes are
almost identical to the Rabi lineshapes calculated for two level systems.10 A
pictorial view of the effect of many cycles is given in Fig. 10.II.12 On
successive cycles the atom samples the level crossing at the turning point of
the field, and the transition amplitudes of many cycles add coherently,
leading to destructive or constructive interference. In other words, there are
resonances, and the Landau-Zener description goes over smoothly to the
resonant photon absorption picture. The other wayof describing the
transitions is based on a Floquet approach.13"16 To illustrate the essential
ideas we use a simple model to calculate the Rabi frequencies for the
problem depicted in Fig. 10.10.There are two states: 1, which has a linear
Stark shift, and 2, which has no Stark shift. They are coupled by the

Microwave multiphoton transitions 175 time independent core coupling V.

We treat the problem byfirstignoring the core coupling and introducing it later
as a perturbation. The unperturbed Hamiltonian, H = Hu + HEs is used to
describe the atom in the static field Es, and the time dependent wavefunctions
of states 1 and 2are given by Eq. (10.4). Adding a microwavefield in the
same direction, Emvv cos cut, corresponding to the Hamiltonian H = Hn +
HEs + HEmw, does not couple states 1 and 2. Thus the wavefunction for
state 1, for example, may be written as ipi(r,t) = �i(0Vi(r)It is a solution of
the time dependent Schroedinger equation .-kEs - kEmw cos a>07\(0Vi(r)
(10.10) which has as a solution for Tx(t) T^t) = Q-^^WI-kEs-kEmw cos cot')
dt'^ (10.11) Integrating Eq. (10.11) yields T1(0=e-i(Wl-*Es)r"^sintttf),
(10.12) where we have dropped the constant phase from the initial time t0.
Finally, Vi(r,r)=Vi(r) e-'fa'-^H, (10.13) where W[ = Wi � kEs. When the
wavefunction is written in thiswayit shows that adding the microwave field
modulates the energy around the value W[ at the angular frequency co. In the
same way that frequency modulation of a radio wave leads to sidebands,
described by a Bessel function expansion, modulation of the atomic
wavefunction also leads to sidebands. Using the fact that17 (x)eina", (10.14)
we can write the wavefunction Vi(r>0 by expressing Eq. (10.12) in terms of
the Bessel function expansion of Eq. (10.14). Explicitly, /� ( ^ ) e^^j ^(r).
(10.15) We can identify W[ as the carrier energy of state 1, and there are
sidebands at energies W[ + nco for all n. There isin principle an infinite
number of terms in the expansion. However, for large argumentsJn(x) ~ 0 if
\n\ > x,17 and asa result, the sideband amplitudes are only significant for
kEmw > \n\co. In other words, in a microwave field the sidebands with
nontrivial amplitude extend as far in energy, �A;�mw, as the energy
modulation produced by the microwave field. In Fig. 10.12 we show the
energies of the sidebands of state 1as a function of the static field. State 2 has
no Stark shift, and thus under the Hamiltonian H = Hn + HEs + HEmw its
wavefunction is given by Eq. (10.4b), and it has no sidebands with nonzero
amplitude. With the Hamiltonian H = HH + HEs + HEmw, i.e. in the

176 Rydberg atoms erg y LLI , 1 \ \ 2 ^ < s I \ \ N \ Reid Fig. 10.12 Energy
level diagram for two states 1and 2, showing the sidebands ( ) of 1 produced
bythe microwave field of angular frequency co (from ref. 12). absence of the
coupling V, the Mh sideband of state 1 is degenerate with and crosses state 2
at ES = EC +^ - (10.16) When the core coupling Vis introduced as a
perturbation, the crossing of state 1 and its sidebands with state 2become
anticrossings, as shown by Fig. 10.12. The magnitude of the avoided
crossing, coN, between the Mh sideband of state 1 and state 2 at the
staticfieldEs = Ec 4- Ncolk is obtained from the time independent part of the
coupling matrix element (ip2 (r,f)Mt/>i(r,0>> . Using the wavefunctions of
Eqs. (10.4b) and (10.15), the time independent part of matrix element is
given by (*p2(r)\V\Mr))N = (2\V\l)JN(kEmJa>). (10.17) Accordingly, =
a)0\JN(kEmJoj)\. (10.18) In essence, the magnitude of the avoided crossing
a)Nis the static field avoided crossing a>0 multiplied by the fractional
amplitude of state 1 in the Nth sideband. The significance of the magnitude of
the avoided crossing isthat it is equal to the frequency at which the atom
makes transitions back and forth between 1and 2, the Rabi frequency. A
familiar example showing the equivalence of the avoided crossing size and
the Rabi frequency isprovided by the analogous anticrossing at Ec between 1
and 2 inapurely staticfield.At Ec the eigenstates are (ip1� ip^lfl. If the two
eigenstates are excited coherently, in a time �1/OJ 0, the population

Microwave multiphoton transitions 111 250 LU 100 200 300 400

MICROWAVE FIELD (V/cm) Fig. 10.13 Total energy (number of photons
times photon energy) absorbed in the K19s-� (17,3) transition vs
microwave field: Experimental points; 9.27 GHz, (A), and 10.353 GHz(�);
theoreticalpoints; Landau-Zener calculations (�) andtwo levelFloquet
calculations (O) (from ref. 10). oscillates between 1 and 2 at frequency co0
=2b, and this oscillation may be detected in a quantum beat experiment.
There are many forms in which the Floquet approach can be more rigorously
implemented.13"16 One which has been used often isthe infinite matrix
approach of Shirley.14 It corresponds roughly to the infinite set of sidebands.
An alternative more compact approach has been described by Sambe15 and
ChristiansenDalsgaard.16 When Landau-Zener and Floquet calculations are
performed to match the K 19S _> (17,3) transitions, with parameters Wt
=220 GHz, W2 =0, k = 400 MHz/ (V/cm), and co0 = 800 Mhz, the Floquet
and multicycle Landau-Zener calculations predict that the same
microwavefieldisrequired to produce a given Rabi frequency for an N
photon transition.10 To compare calculated transition strengths to the
experimental observations requires that we decide what is an observable
Rabi frequency. In Fig. 10.9 the observed resonances are ~1 GHz wide while
the exposure time of the atoms to the microwaves is 1 JUS. Making the
questionable assumption of homogeneous broadening leads to a required
Rabi frequency of �30MHz, and this Rabi frequency was chosen to compare
to the experimental results. While the calculations agree with each other, they
do not exactly match the experiments, as shown by Fig. 10.13, a plot of the
total energy absorbed from the 9 or 10 GHz microwave field as a function of
microwave field strength. In Fig. 10.13, absorbing 150 GHzcorresponds
tofifteen10 GHz photons

178 Rydberg atoms or seventeen 9 GHz photons. The calculated fields

required to drive the n photon transitions are systematically 25% higher than
the experimental results, a difference which is probably due to
oversimplifications in the model. For example, the dipole coupling between
the levels, is ignored. It is interesting to note that Fig. 10.13 also
demonstrates graphically the fact that the number of photons absorbed
increases linearly with the microwave field amplitude. Most important,
however, it isapparent that the multiphoton resonances can be described
equallywell intwo different ways. We have thus far treated the K (n + 2)s
states as having no Stark shifts. While they have no first order Stark shift,
they do have a second order shift due to their dipole interaction with the p
states, which are removed from the s states by energies large compared to the
microwave frequency. The microwave field does not produce appreciable
sidebands of the s state since it has no first order Stark shift. However, it
does induce a Stark shift to lower energy. Not surprisingly the Stark shift
produced by a low frequency microwave field of amplitude E is the same
asthe second order Stark shift produced by astaticfieldEs/j2, they have the
same value of (E2). Careful inspection of Fig. 10.9 reveals that the
resonances observed with high microwave powers shift with power. In fact,
the second order Stark shift of the (n + 2)s state due to the microwave
fieldcan bring the (n + 2)s�> (n,k) transitions into resonance with no static
field. This Stark shift is the explanation of the anomalous K 19s �> (17,k)
threshold shown in Fig. 10.8. The transitions at microwavefieldsfrom 500to
600V/cm, i.e. fields below the avoided crossing field, Ec = 546 V/cm, are
due to the fact that the 19s state isshifted into 27 photon 19s-(17,A:)
resonances in thisfieldrange, and the Rabi frequencies of the transitions to a
few, three or four, of the (11,k) states, k ~ 3, are adequate to observe the
transitions in thisfield range. At higher fields, �700 V/cm, the transition
becomes nonresonant, and is not observed again until E ~ 750V/cm,
corresponding to the 28 photon resonances. At this field, the Rabi
frequencies of many more 19s �> (17,/:) transitions are adequate to allow
observation of the resonances, and what is observed as a threshold is really
many overlapping resonant transitions. In retrospect, it is now clear why the
threshold fields for the (n + 2)s�> (n,k) transitions observed with
microwaves alone are scattered from 1.0�c to 1.6EC, contrary to the
prediction of a single cycle Landau-Zener model. Both the resonance
condition and Rabi frequency condition must be fulfilled, and thefieldat
which the resonance condition is fulfilled is rather random. Having
considered the connection between the multiphoton resonances and the
microwave threshold field for the K (n + 2)s �> (n,k) transitions, it is now
interesting to return to the analogous n�> n + 1 transitions which are
responsible for microwave ionization and consider them from this point of
view. We start with a two level description based on the extreme n and n + 1
\m\ = 0 Stark states, a description which is the multiphoton resonance
counterpart to the single cycle Landau-Zener model presented earlier. The
problem is identical to the problem

Microwave multiphoton transitions 179 depicted in Fig. 10.10 except that

both states have Stark shifts and sidebands. For simplicity we label these
two extreme Stark states as n and n + 1. The Hamiltonian for this problem is
identical to that of Eq. (10.2) except that there is no static field, i.e. H = Hn +
HEmw + V. UsingH = Hu + HEmwwe find the solutions to Eq. (10.2)
analogous to the solution given in Eq. (10.15), i.e.1617 V Jkfc^M e-*��
(10.19a) V l w I and (r,0 = VWi(r) e*^1? J / J ^ M *-*'�*. (10.19b) � \ CO
) We wish to calculate the Rabi frequency for the TVphoton transition at
resonance, i.e. Nco = l/2n2 � l/2(n + I)2 ~ 1/n3. The matrix element of V
coupling the two states of Eqs. (10.19) is given by = (n + \\V\n)<? J 0) k k
(10.20) At the A^ photon resonance, when Nco = 1/n3, the n and n + 1 states
are degenerate and are coupled by the time independent part of the coupling
matrix element of Eq. (10.20), which is Using the relation17 JN(X ~ y) = �
h-N(x)h(y) (10.22) k Eq. (10.21) may be written as l ( k k ) E A (10.23) Eq.
(10.23) can be applied to any pair of n and n + 1Stark states. If we consider
the extreme m = 0 states, for which A:n = 3rc(n - l)/2 and fcn+1 = - 3rc(rc +
2)/2, Eq. (10.23) reduces to (^A (10.24) Eq. (10.23) implies that it is only
the difference in the Stark shifts which is important.1017 Again the fact that
for large TV, /N(x) �> 0 for Af > x leads to the conclusion that �mw >
l/3n5 to have anonnegligible coupling of the extreme Stark

180 Rydberg atoms ts ) (arb .un i GNA L 0 0 11 i 2 0 0.01 xio"4 dyT\ n*4E
(a.u.) 0.015 / 0.02 i i 1 1 0 50 75 100 125 150 175 200 E (V/cm) Fig.
10.14Ionization signal ofHe 28s3S atoms asa function ofthepeak electricfield
inside the cavity. The size of the vertical scale corresponds toapproximately
1/3 of the saturated signal. Inset: the calculated transition probability from
the28s3S to the29s3S state after one field cycle (from ref.3). states. Slightly
higher fields are required to have significant coupling between states other
than the extreme ones. Of course the resonance condition must also be met.
For ionization by a fixed frequency microwave field tuning must come from
the AC stark shifts, which are sufficient. If only the extreme Stark states
participated in the transitions, sharp resonances might be observed.
However, in the n �> n + 1 transition there are ~n possible initial states in
the n Stark manifold as well as n + 1 final states in the n + 1Stark manifold,
leading to n(n + 1) transitions in a tuning range of co. For a frequency of 10
GHz at n = 20there is a resonance every 25 MHz on the average. When the
microwave field reaches l/3n5 some of these resonances begin to have
observable Rabi frequencies, and the onset of these closely spaced, usually
overlapping, resonances produce what appears to be a threshold field for the
transition. Since the resonance criterion is not an issue, the primary criterion
is that the Rabi frequency be high enough, leading to the observed E ~ l/3n5
dependence. Due to the n(n + 1) possible n->n + 1 transitions it is in general
difficult to observe resonance effects in microwave ionization as obvious as
those shown in Fig. 10.9. Nonetheless several experiments show clearly the
importance of multiphoton resonance in microwave ionization. In Ba and in
He the observed microwave ionization thresholds are structured by
resonances3'6. An excellent example is the microwave ionization probability
of the He 28 3 S state shown in Fig. 10.14. In He the 3 S states intersect the
Stark manifold at fields approaching l/3/i5, and as a result making transitions
from the energetically isolated 3 S state requires afieldcomparable to
thefieldrequired to drive n-^n + 1 transition. The structure in Fig. 10.14 is
quite similar to the structure in Fig. 10.8, which is not

Microwave multiphoton transitions 181 100 200 300 E (V/cm) Fig. 10.15
Observed Na 3p �> n = 25,26 spectra in varying strengths of microwave
field overlaid on anenergy level diagram of theNa \m\=0 states.Thebaseline
of eachspectrum is located at the amplitude of the microwave field. Note that
only odd sidebands of the p states occur and that the predominant effect of
higher microwave fields is to add more sidebands (from ref. 18). surprising
since most of tuning comes from the second order Stark shifts of the s states
in both cases. In alkali atom experiments no explicit resonances have been
observed in microwaveionization. However, there areindirect confirmations
of themultiphoton resonance picture. First, according to the multiphoton
picture the sidebands of the extreme n and n 4- 1Stark levels should overlap
if E = l/3n5. In the laser excitation spectrum of Na Rydberg states from the
3p3/2state in the presence of a 15 GHz microwave field van Linden van den
Heuvell etal. observed sidebands spaced by 15.4 GHz, as shown in Fig.
10.15.18 The extent of the sidebands increases linearly with the
microwavefield,asshown inFig. 10.15, and the n = 25 andn =26sidebands
overlap at microwavefieldsof 150V/cmorhigher, matching the observation
that the 25d state has an ionization threshold of 150 V/cm in a15 GHz field.
Pillet et al. observed that adding small static fields dramatically reduces the
microwavefieldsrequired for the ionization of Li.19 For example the
application of astaticfieldof 1 V/cmlowers the 15 GHzionization threshold
ofthe Li42d state from �200 V/cm, to abroad threshold centered at 20V/cm,
afieldonly slightlyin excess of E = l/3n5 = 13V/cm. The threshold field 200
V/cm corresponds to the hydrogenic thresholdfieldof 1/9 n4, which will be
described shortly. A small field has virtually no effect in a single cycle
Landau-Zener model, but its dramatic
182 Rydberg atoms effect is readily explained in a multilevel resonance
picture. The small static field lifts the degeneracy of sidebands from different
Stark states, transforming them into a quasi-continuum of states, ensuring that
the resonance condition is met. Finally, Mahon etal.5 have observed that
even at frequencies aslow as670 MHz that ionization of Na occurs at fields
near l/3/i5. At such low frequencies it is impossible to explain the ionization
on the basis of incoherent single cycle Landau-Zener transitions. Rather the
coherent effect of many cycles of thefieldis required. Hydrogen Microwave
ionization of H and H like atoms differs radically from the ionization of
nonhydrogenic atoms. For our present purposes, an atomic system isH like if
it has small quantum defects. For example the Na \m\ = 2 states are composed
of \m\ >2 states all of which have quantum defects <1.5 x 10~2. With small
quantum defects the avoided crossings between the Stark levels of adjacent n
are negligibly small, and there is negligible coupling between n levels due to
the core interaction. As this coupling is responsible for the ionization of
nonhydrogenic atoms at E = l/3n5, it is not surprising that H like atoms do not
exhibit ionization at E = l/3n5. It is worth noting that atoms may appear to be
hydrogenic in one case but not in another. For example, the Na \m\ = 2states
are ionized by 15GHz fieldsof l/9n4, as shown by Fig. 10.3, but can be
ionized at lower fields, ~l/3n5, when a static field is applied to produce a
quasi-continuum of levels.20 Ionization at E = l/3n5 does not occur in H
under the same conditions. Thus, even quantum defects of 10~2 are adequate
to produce nonhydrogenic behavior in some circumstances. The first
measurements of microwave ionization in any atom were carried out with a
fast beam of H by Bayfield and Koch1, who investigated the ionization of a
band of approximately five n states centered at n = 65. Using microwave and
rf fields with frequencies of 9.9 GHz, 1.5 GHz, and 30 MHz, to ionize the
atoms they found that the samefieldwasrequired at 30 MHz and 1.5 GHz
toionize the atoms, but that a smaller field was required at 9.9 GHz. The
measurements showed that at n = 65 frequencies up to 1.5 GHz are identical
to a static field. Later, more systematic measurements have confirmed the
initial measurements and have allowed significant refinements of our
understanding. In Fig. 10.16 we show the ionization threshold fields (in this
case the field at which there is 10% ionization) of H in a 9.9 GHz field.21
The ionizationfieldsare plotted as n4E vsn3a), and they bring out two factors.
First, at low frequencies the field required is ~l/9n4, the staticfieldrequired
to ionize the red n Stark state of \m\ � n. Second, asshownby the scaling of
the horizontal axis, the required field drops below l/9n4 as the microwave
frequency approaches the interval between adjacent n states, 1/n3.

Hydrogen 183 0.10 "c o o CO " 0.05 LU c 5 � X 5 0 ~ X \v > _ 1 1 1 1 1 -

DX O i I I i experiment 1-dim theory 2-dim theory i i �0 0.2 0.4 0.6
n3cu(atomic units) Fig. 10.16 Scaled microwave ionization field, n4E, for H
plotted against the scaled microwave frequency n3a>: experimental (�); one
dimensional theory (x);twodimensional theory (O). n3co = 0.05 corresponds
to n = 32and n3co = 0.6 corresponds to n = 73. Note the declinefrom n4E =
1/9 at n = 30 toprogressively lowervaluesas napproaches 60 (from ref. 21).
Qualitatively similar observations have been made for co < l/3n3 for the H
like Na and Li \m\ = 2 states with 15 GHz fields.22 When co � 1/n 3 , the
field required for ionization is E=l/9n4, and as co approaches 1/n3 it falls to
E^O.OArT4. These observations can be explained qualitatively in the
following way. At low n, so that co � 1/n 3 , the microwave field induces
transitions between the Stark states of the same n and m by means of the
second order Stark effect. With only a first order Stark shift a state always
has the same dipole moment and wavefunction, as indicated by the constant
slope dW/dE of the energy level curve. Thus when the field reverses, E�>
� E, the Rydberg electron's orbit does not change. With a second order Stark
shift aswell, the slope dW/dE is not the same at E and �E, and as a result
the dipole moment and wavefunction are not the same. If the field isreversed
suddenly asingle Stark state in the field E is projected onto several Stark
states of the same n and m when E-> � E. Since all the Stark states of the
same n make transitions among themselves they ionize once the field is
adequate to ionize one of them, the red one, at E = l/9n4 for \m\ � n. As co
�> 1/n3, or more commonly, as n is increased with fixed co,the required
field falls below l/9nA as transitions to higher n occur, allowing ionization at
lower fields. Since many energy levels are coupled this development has
been explained

184 Rydberg atoms as the onset of classical chaos.23 It can also be

explained in terms of a rate limiting step. We here present a development
along the latter line, due to Christiansen-Dalsgaard,24 to calculate the
hydrogenic microwave ionization fields for a> approaching but not
exceeding 1/n3.We assume that ionization is always possible ifE > l/9n4,
asoutlined above. On the other hand asCD �> 1/n3 the fieldrequired to drive
the rate limiting n �� n + 1transition falls below l/9n4, and ionization
occurs by transitions tohigher n states followed by ionization from the higher
n states. The central problem is to calculate the field required to drive the n-
*n + 1 transition via an electric dipole transition. In the presence of an
electric field, static ormicrowave, the natural states touse are the parabolic
Stark states. While there is no selection rule as strict as the A�= � 1
selection rule for angular momentum eigenstates, it is in general true that each
n Stark state has strong dipole matrix elements to only the one or two n + 1
Stark states which have approximately the same first order Stark shifts. Red
states are coupled to red states, and blue to blue. Explicit expressions for
these matrix elements between parabolic states have been worked out,25
and, as pointed out by Bardsley etal26, the largest matrix elements are those
between the extreme red or blue Stark states. These matrix elements aregiven
by(n \z\n + 1) = n2/3.26 Since the extreme n and n + 1 Stark states have the
largest coupling matrix elements, it seems reasonable to assume that the
n�>n + 1 transitions occur through these states and calculate the field
required to drive the transition between this pair of levels. While this is an
approximation, it is useful, just as considering the extreme Stark states gives
a reasonable description of the ionization of Na. Following theapproach of
Christiansen-Dalsgaard,24 we can calculate the n �� n + 1Rabi frequency.
We write theHamiltonian as H = Hu +HEmw +HAnEmw. (10.25) HH has the
same definition as before, HEmw includes only the diagonal matrix elements,
and HAnEmw includes theoff diagonal elements. H = HH + HEmw has as its
eigenfunctions the linear hydrogenic Stark states. The dominant effect of
HAnEmw is the dipole coupling between n states. If we use H = Hu +HEmw
the wavefunctions for then and n + 1 blue Stark states aregiven by Eq.
(10.19) with kn = 3n(n - l)/2 and kn+1 = 3n(n + l)/2. At the n photon
resonance Nco = l/2n2 � l/2(n + I)2 = 1/n3, and we can calculate the Rabi
frequency starting from the dipole coupling matrix element,24'27
(ipn+1(r,t)\HAnEJipn(r, 0) = (i/>n+1(r,t)\zE cos (ot\yn(r,t)). (10.26) Using
the Bessel function forms of the wavefunctions given by Eqs. (10.19), writing
coscot in exponential form and using Eqs. (10.19), we identify the time
independent part of Eq. (10.26) as
Hydrogen 185 2 \Y \ co IV co 11 � mw\ , /, Em, + Jk-l-N \kn+l \Jk\kn \ CO )
\ CO Using Eq. (10.22), Eq. (10.27) becomes IJN+A� 1 ^\ co CO ,17
(10.28) \ co )) Using1^ -/v (x) (10.29) X allows us to write Eq. (10.28) as j
JN\ \ co For the extreme Stark states (n+l|z|w) = n2/3, and kn+1 � kn = 3n.
At the photon resonance Nco = 1/n3, and the N photon matrix element is
given by co I and the Rabi frequency is twice as large. In this form it is clear
that it is predominantly the variation of the Bessel function which determines
the field required to achieve a useful Rabi frequency. If we assume that many
photons are required for the n�> n + 1 transition, N is large and JN (3nE/co)
�> 0 unless 3nElco > N. Since Nco = 1/n3, this requirement can be restated
as E>� A, (10.32) a field above the static ionization field of E = l/9n4.
Evidently, for co � 1/n 3 ionization does not occur by transitions to higher
states but by direct field ionization as described earlier. On the other hand
when N is a small number, <5, the above reasoning for large arguments of the
Bessel function does not apply. Recall that in the K (n 4-2)s�> (n,k)
transitions the first few transitions require very small fields. Similarly, when
n < 5, the Rabi frequency can be appreciable even for E � l/3n 4 . Explicit
evaluation of Eq. (10.31) yields immediately the Rabi frequency for the
transitions. An upper limit to the required Rabi frequency is found by
requiring that it be equal to the detuning. Using this requirement with Eq.
(10.31) Christiansen-

186 Rydberg atoms 10' Fig. 10.17 The criticalfield,forionization, E, as

function ofthe principal quantum number nofthe lower
statefordipoletransitionsinH ( ).Experimental pointsfor Na (�) andLi (O)
areshown as well asthe criticalfieldfor tunneling ( ) (from ref24). Dalsgaard
has computed the hydrogenic fields for ionization by a 15 GHz field.24 The
results of this two state model are compared to the experimental Na and Li
m\ = 2thresholdfieldsin Fig. 10.17.The computed thresholdfieldsare in
general agreement with the observations, although they exhibit more structure
than do the experimental points. For example the two photon resonance at n =
60 is apparent in the theoretical curve but not the experimental points. The
discrepancies are due to the simplicity of the model. In the theoretical model
the second order Stark shifts and the existence of other stateswere ignored.
Onlythe extreme n and n + 1 Stark states were considered, but just asin the
Na or Kcases the other Stark states also make n�� n + 1 transitions
although, usually at slightly higher fields due to their smaller n-� n +
1matrix elements. However, the fact that the carrier energies of different
Stark states of the same n are different, due to their different second order
Stark shifts, means that it is not obvious which transitions between n and n +
1 Stark states are more likely to be resonant. When the many n and n 4- 1
Stark states and possible variations of the microwavefieldwith time are taken
into account, it is apparent that sharp resonances are less likely to be as
apparent as they are in the two state theoretical curve of Fig. 10.17. If the
microwave ionization can be described by a resonance multiphoton picture,
it should be possible to observe other manifestations of resonance

Hydrogen 187 0.0 o.io Fig. 10.18 Ionization probability Px of H n =36 atoms
by 9.92 GHz field experimental (curve (e)), one dimensional adiabatic
quantum calculation (curve (1)), one dimensional classical calculation (�),
three dimensional classical calculation (O)(from ref. 28). phenomena, and
such is the case. Ionization thresholds in H show, in some cases, marked
structure. An example of this phenomenon is shown in Fig. 10.18, the
structured ionization curves of H n = 36atoms in a9.92 GHz field.28 The
origin of the pedestal at E ~ 0.12 n~4 is due, in the terms used above, to a
resonance. It is similar to the K and He resonances shown in Figs. 10.8 and
10.14. The ionization curve of Fig. 10.18 isobtained in the same way asthe
data shown in Fig. 10.14, by exciting atoms in zero field and then exposing
them to a strong microwave field. When atoms are excited in the presence of
a static field, to a single Stark state, and held in single Stark state bythe
continued application of the field, resonances became more apparent when a
microwave field in the same direction is applied. Bayfield and Pinnaduwage
have observed transitions from the extreme red H n = 60, m = 0 Stark state to
other nearby extreme Stark states in static fields of 5-10 V/cm.29 As shown
by Fig. 10.19 resonances corresponding to the four photon transition to the
extreme red n = 61 Stark state and four and five photon transitions to the
extreme red n = 59 Stark state are visible. These experiments are similar to
the K and He multiphoton resonance experiments described earlier, but are
inherently simpler because the extreme red n = 60 Stark state is only coupled
to the extreme n = 59Stark state. In contrast, the K (n + 2)s state is coupled to
all the (n,k) Stark states. As shown by Fig. 10.16, as the microwave
frequency approaches I//?3 the scaled field n4E required for ionization drops
steadily, raising the question of what
188 Rydberg atoms n-changing up, 0.2 watt tli I f i* 6.0 6.5 7.0 7.5 8.0
Microwave Frequency (GHz) Fig. 10.19 The microwave frequency
dependence of the n changing signals at low microwave power, where n
changes up or down only by 1. Resonant multiphoton transitions areobserved
near theexpected static field Stark shifted frequencies indicated. These
resonances involve the absorption of four or five microwave photons.
Thedown n changing atom production curve was obtained with the state
analyzer field EA set at 50.0V/cm, while up nchanging wasstudied asn = 60
atom loss with EA =45.5 V/cm. The locationsof resonancesfor larger direct
(not stepwise) changesin nareindicated alongwith their order k (from ref.
29). happens when co exceeds 1/n3. Reaching the regime co > 1/n3 requires
higher n, higher frequency, or both. For example, at 10 GHz co = IIn3 at n =
87. Galvez et al.30 have observed the ionization of H n = 40 to n = 80 states
by a 36 GHz microwave field. They took data in two modes; the quench
mode, observing the atoms remaining in the initial state, and the ionization
mode, observing those which have been ionized. Static fields downstream
from the microwave field region were used to analyze the final states, and in
the ionization mode the microwave ionization signal also included bound
states of n > nc, where 160 < nc < 190. Their results, obtained in the
ionization mode are shown in Fig. 10.20. As shown by Fig. 10.20, when co
exceeds 1/n3the scaled field, n4E, required to ionize H no longer drops
sharply but in fact increases very slightly with n. At 36 GHz n3co = 0.4 and
2.8 correspond to n = 42 and 80. The measured ionization fieldsare not
inconsistent with a constant value of n4E for n3co > 1. Jensen et al.31 have
developed a simple quantum mechanical model for the regime, IIn3 � co �
1/n 2 which predicts an n independent ionization field. In their model
ionization occurs by a sequence of single photon transitions through higher
lying

Hydrogen 189 0.08�*�� 0.8 1.2 1.6 2.0 2.4 2.8 n03u (scaled freq. for
<J/2 7T = 36.02 GHz) Fig. 10.20 H36 GHzmicrowave ionization fields
(10% ionization):experimental ionization mode results (o); three dimensional
classical calculations (x, ()) (from ref. 30). states. The sequence of
transitions, and ionization, only occurs if the Rabi frequencies of the
transitions equal or exceed the detunings. If the energy difference between
two states of principal quantum numbers n and n' is given by II = H2n'2 �
l/2n2, the electric dipole matrix element connecting these two states is given
by31 (n\z\nf) = 0A(nnf)-3/2^~5/3 = 0An-3n~5/3 (10.33) provided that n,n' �
1and \n � n'\ � n. If the two requirements for Eq. (10.33) are fulfilled, the
dipole matrix element for the n�>n' transition scales as n~3, just as
separation between n states and the detuning from resonance. As a result,
requiring that the Rabi frequency equal the maximum detuning, ~l/2n3 leads
to the requirement E = 2.5fl5/3. (10.34) Since resonance fl = co, the
microwave frequency, the predicted ionizing field is n independent. For a
microwave frequency of 36 GHz the field predicted by Eq. (10.34) is 10
V/cm, in reasonable agreement with Fig. 10.17. As co increases further
above 1/n3 a single photon drives the initially populated state closer and
closer to the ionization limit, and ionization occurs with the absorption of
fewer photons. Few photon processes are well described by lowest order
perturbation theory, which shows that the rates are proportional to E2N,
where N is the number of photons absorbed. For small iVsuch processes are
not well described by a threshold field, and it is not meaningful to discuss
ionization threshold fields in this case. The discussion of microwave
multiphoton processes given here is largely a Floquet description, i.e. a
steady state description. We have implicitly assumed,

190 Rydberg atoms for example, that the microwavefieldsare turned on and
off rapidly compared to the Rabi frequencies and that the time during which
the microwaves are turned on or off is so short that it contributes nothing to
the transition. It is clear that these assumptions are not always met, and
interesting dynamic effects due to tuning through resonances by second order
Stark shifts have been predicted and observed.32"34 Similarly, we have not
considered the effects of noise as anything other than a source of line
broadening. In fact noise can have a profound effect, especially in a
broadband microwave system where broadband noise almost ensures
coincidences with relevant atomic transitions.35 Ionization bycircularly
polarized fields While ionization by linearly polarized fields has been well
studied, there is only one report of ionization by a circularly polarized field,
the ionization of Na by an 8.5 GHz field.36 In the experiment Na atoms in an
atomic beam pass through a Fabry-Perot microwave cavity, where they are
excited to a Rydberg state using two pulsed tunable dye lasers tuned to the 3s
�� 3p and 3p�> Rydberg transitions at 5890 A and �4140 A
respectively. The atoms are excited to the Rydberg states in the presence of
the circularly polarized microwavefieldwhich is turned off 1 JUS after the
laser pulses. Immediately afterwards a pulsedfieldisapplied tothe atoms to
drive any ions produced by microwave ionization to a microchannel plate
detector. To measure the ionization thresholdfield the ion current ismeasured
as the microwave power is varied. The experimental approach toproducing
the circularly polarizedfieldisto usea Fabry-Perot cavity which supports two
orthogonally linearly polarized 8.5 GHz fields 90� out of phase. The major
experimental problem is to minimize the interaction between these two
nominally degenerate cavity modes, as any interaction lifts the degeneracy.
To minimize the interaction Fu et al. fed the two modes from orthogonally
polarized waveguides through irises in the mirrors.36 The two modes were
offset by 2 MHz, and they had Qs of 2000 and 2100, so the linewidths of the
modeswere 4MHz. Due to the fact that the Qs were not identical circular
polarization could exist in steady state, but not as the cavities modes were
filled with the microwave field. To minimize the exposure of the atoms to
elliptically polarized microwave fields Fu et al. excited the atoms to a
Rydberg state in the microwave field. A much higher circularly polarized
field is required to ionize the atoms than a linearly polarized field, as shown
by Fig. 10.21, a plot of the threshold fields, where 50% ionization occurs,
for linearly and circularly polarized 8.5 GHz fields. As shown by Fig. 10.21,
the circularly polarized microwave ionization threshold fieldis very nearly E
= l/16n4, the same as the the staticfieldrequired to ionize a Rydberg Na atom
and much higher than the field required for ionization by

Ionization by circularly polarized fields 191 LU 101 3 4 5 106 2 3 4 5 n4

Fig. 10.21 Ionization threshold fields for linear (�) and circular (�)
polarization as a function of nwhen Na atoms are excited in the 8.5 GHz
microwavefield(from ref. 36). linearly polarized microwaves, E = l/3n5.
Furthermore there is a sharp dependence of the observed ionization signal on
the relative phase between the two polarizations in the cavity. When E is
20% below l/16n4, so that ionization by a circularly polarized field does not
occur, a 10� phase deviation from the 90�phase shift required to produce
circular polarization leads to nearly complete ionization. At the same
microwave power, a small ellipticity in the polarization produces ionization,
while pure circular polarization does not. If we transform the problem to a
frame rotating with the microwave field, it is static and cannot induce
transitions. The transformation to the rotating frame, often used to describe
two level magnetic resonance experiments, is discussed by Salwen37 and
Rabi et al.38. If we have a microwave field given by E = xE cos cot + yE sin
cot, (10.35) and we transform the problem to a frame rotating about the z axis
at angular frequency co, thefieldbecomes astaticfieldin thex direction inthe
rotating frame. If we use as eigenstates in the laboratory frame the usual
spherical n(,m states, a

192 Rydberg atoms rotation through angle 0 about the z axis transforms the
wave function ipntm to zim<t>ipnzm. For uniform rotation <p = cot and t/Wm
~*elmft^VWm>ie- >tne energy is shifted by �mo>. Transforming the Na
n�m wave functions to the rotating frame only adds �mco to their energies.
In principle, it is then a simple matter to diagonalize the newHamiltonian
matrix containing afieldin thex direction to find the energy levels in the
rotating frame. In practice, the fact that there are n2l2 coupled levels for each
principal quantum number complicates the problem. To develop an
understanding Fu et al. diagonalized the Na Hamiltonian matrix, for n � 5, 6,
and 7; in a frame rotating at 1500 GHz, 2% of the n = 4-5 interval, to
correspond to the fact that 8.5 GHz is 2% of the n = 25 to 26 interval. When
the Stark manifolds of adjacent n overlapthere are avoided crossings, just as
in astatic field. In the limit of very low frequency the Stark manifolds overlap
at E = l/3n5, but the overlap occurs at lower fields as the frequency is raised.
In analogous calculations for H the levels cross, as in a static field. Thus the
avoided crossings in Na are due to the core coupling not the transformation
to the rotating frame. In static fields l/3n5 < E < l/16n4 the core coupling is
manifested in avoided level crossings, and in fields E > l/16n4 the same
coupling between discrete states and the underlying Stark continuum of
ionized red Stark states leads to ionization.39 In other words field ionization
of Na at E > l/16n4 is really autoionization, and in the rotating frame the
situation is presumably the same. There are avoided crossings for E <
l/16rc4, presumably due to the m = 0 and � 1 parts of the wavefunction, and
ionization when E > 1/16/t4. In other words, ionization by a circularly
polarized microwave field is field ionization in the rotating frame. The
pronounced effect of ellipticity in the polarization can also be understood
with the rotating frame description. When the polarization is elliptical, in the
rotating frame there is a field oscillating at 2co superimposed on the static
field. If the staticfieldin the rotating frame exceeds l/3n5, the atom is
inthefieldregimein which there are many level crossings, and even a very
small additional oscillating field can drive transitions to higher lying states
via these level crossings, leading ultimately to ionization. Nauenberg40 has
approached the problem of ionization by acircularly polarized field
classically, by transforming the problem to a rotating frame. In the rotating
frame the potential is depressed bythe addition of aterm �co2(x2 + y2)/2. At
lown the ionization field is given by E = l/16n4, but at high n the required
ionization field falls below H16n4. As an example, at 8.5 GHz for n = 30, E
= l/16n4, but for n = 50the classically calculatedfield isabout afactor
offivebelow E � l/16n4. As shown by Fig. 10.21, the experimental results
exhibit a l/16n4 dependence up to n = 60. While ionization of n = 50atoms is
energetically allowed at a much lower field, it does not happen. A possible
impediment is an angular momentum constraint.41 At the threshold for
ionization the electron barely escapes over the saddle point in the rotating
frame. In the laboratory frame the same electron hasa large angular
momentum. For example for n = 50it has an angular momentum of

References 193 60ft. In quantum mechanical terms acquiring this much

angular momentum requires ten Arc= 1 transitions, which is unlikely.
References 1. J. E. Bayfield and P. M. Koch, Phys.Rev. Lett. 33, 258, (1974).
2. P. Pillet, W. W. Smith, R. Kachru, N. H. Tran, and T. F. Gallagher,
Phys.Rev. Lett. 50, 1042 (1988). 3. D. R. Mariani, W. van de Water, P. M.
Koch, and T. Bergeman, Phys.Rev. Lett.50, 1261 (1983). 4. P. Pillet, H. B.
van Linden van den Heuvell, W. W. Smith, R. Kachru, N. H. Tran, and T. F.
Gallagher, Phys.Rev. A 30, 280 (1984). 5. C. R. Mahon, J. L. Dexter, P.
Pillet, and T. F. Gallagher, Phys.Rev.A 44, 1859 (1991). 6. U. Eichmann, J.
L. Dexter, E. Y. Xu, and T. F. Gallagher, Z. Phys.D 11,187 (1989). 7. J.
Rubbmark, M. M. Kash, M. G. Littman, and D. Kleppner, Phys.Rev. 23, 3107
(1981). 8. L. A. Bloomfield, R. C. Stoneman, and T. F. Gallagher, Phys.Rev.
Lett SI, 2512(1986). 9. R. C. Stoneman, G. R. Janik, and T. F. Gallagher,
Phys.Rev. A 34,2952 (1986). 10. R. C. Stoneman, D. S. Thomson, and T. F.
Gallagher, Phys.Rev. A 37,1527 (1988). 11. W. van de Water, S. Yoakum, T.
van Leeuwen, B. E. Sauer, L. Moorman, E. J. Galvez, D. R. Mariani and P.
M. Koch, Phys.Rev. A 42, 872 (1990). 12. T. F. Gallagher, inAtoms inIntense
Laser Fields, ed. M. Gavrila (Academic Press, Cambridge, 1992). 13. S. H.
Autler, and C. H. Townes, Phys.Rev. 100,703 (1955). 14. J. Shirley,
Phys.Rev. 138, B979(9165). 15. H. Sambe, Phys.Rev. A 7, 2203 (1973). 16.
B. Christiansen-Dalsgaard, unpublished, (1990). 17. M. Abramowitz and
LA. Stegun, Handbookof Mathematical Functions, (U.S. GPO, Washington,
DC, (1964). 18. H. B. van Linden van den Heuvell, R. Kachru, N. H. Tran,
and T. F. Gallagher, Phys. Rev. Lett. 53, 1901 (1984). 19. P. Pillet, C. R.
Mahon, and T. F. Gallagher, Phys. Rev. Lett.60,21 (1988). 20. G. A. Ruff and
K. M. Dietrick (unpublished). 21. K. A. H. van Leeuwen, G. V. Oppen, S.
Renwick, J. B. Bowlin, P. M. Koch, R. V. Jensen, O. Rath, D. Richards, and
J. G. Leopold, Phys.Rev. Lett. 55,2231 (1985). 22. T. F. Gallagher, C. R.
Mahon, P. Pillet, P. Fu and J. B. Newman, Phys.Rev.A 39, 4545 (1989). 23.
R. V. Jensen, S. M. Susskind, and M. M. Sanders, Phys. Rept. 201, 1 (1991).
24. B. Christiansen-Dalsgaard (unpublished). 25. H. A. Bethe and E. A.
Salpeter, Quantum Mechanics of One and Two Electron Atoms (Academic
Press, New York, 1957). 26. J. N. Bardsley, B. Sundaram, L. A.
Pinnaduwage, and J. E. Bayfield, Phys.Rev. Lett. 56,1007 (1986). 27. M. A.
Kmetic and W. J. Meath, Phys.Lett. 108A, 340 (1985). 28. D. Richards, J. G.
Leopold, P. M. Koch, E. J. Galvez, K. A. H. van Leeuwen, L. Moorman, B.
E. Sauer and R. V. Jensen, /. Phys. B. 22, 1307 (1989). 29. J. E. Bayfield and
L. A. Pinnaduwage, Phys.Rev. Lett. 54, 313 (1985). 30. E.J. Galvez, B. E.
Sauer, L. Moorman, P. M. Koch, and D. Richards, Phys.Rev. Lett. 61, 2011
(1988). 31. R. V. Jensen, S. M. Susskind, and M. M. Sanders, Phys.Rev. Lett.
62, 1476 (1989). 32. H. P. Breuer, K. Dietz, and M. Holthaus, Z. Phys D 10,
13 (1988). 33. M. C. Baruch and T. F. Gallagher, Phys.Rev. Lett. 68,3515
(1992). 34. S. Yoakum, L. Sirko, and P. M. Koch, Bull.Am.
Phys.Soc.37,1105 (1992).

194 Rydberg atoms 35. R. Blumel, R. Graham, L. Sirko,U. Smilansky,H.

Walther, and K. Yamada, Phys.Rev. Lett. 62, 341 (1987). 36. P. Fu, T. J.
Scholz, J. M. Hettema, and T. F. Gallagher, Phys.Rev. Lett. 64,511 (1990).
37. H. Salwen, Phys.Rev. 99, 1274 (1955). 38. 1.1. Rabi, N. F. Ramsey, and
J. Schwinger,Rev. Mod. Phys.26,167 (1955). 39. M. G. Littman, M. M. Kash,
and D. Kleppner, Phys.Rev. Lett.41,103 (1989). 40. M. Nauenberg,
Phys.Rev. Lett. 64,2731 (1990). 41. T. F. Gallagher, Mod. Phys.Lett. B 5,239
(1991).

11 Collisions with neutral atoms and molecules The large size and low
binding energies, scaling as n4 and n 2 , of Rydberg atoms make them nearly
irresistible subjects for collision experiments. While one might expect
collision cross sections to be enormous, by and large they are not. In fact,
Rydberg atoms are quite transparent to most collision partners. Collisions
involving Rydberg atoms can be broken into two general categories,
collisions in which the collision partner, or perturber, interacts with the
Rydberg atom as a whole, and those in which the perturber interacts
separately with the ioniccore and the Rydberg electron. The difference
between these two categories is in essence a question of the range of the
interaction between the perturber and the Rydberg atom relative to the size of
the Rydberg atom. A few examples serve to clarify this point. A Rydberg
atom interacting with a charged particle is a charge-dipole interaction with a
1/R2 interaction potential, and the resonant dipole-dipole interaction
between two Rydberg atoms has a 1/R3 interaction potential. Here R is the
internuclear separation of the Rydberg atom and the perturber. In both of
these interactions the perturber interacts with the Rydberg atom as a whole.
On the other hand when a Rydberg atom interacts with a N2 molecule the
longest range atom-molecule interaction isadipole-induced dipole interaction
with a potential varying as 1/R6. This interaction is vanishingly small and
negligible for R greater than the orbital radius of the Rydberg atom. As a
consequence, only when the N2 molecule actually penetrates the Rydberg
electron's orbit is there any appreciable interaction. Once the N2 molecule is
inside the Rydberg electron's orbit itcaninteract separatelywiththecharged
ionic core and the Rydberg electron. In this chapter the focus isprimarily on
collisions of this type. The discussion of collisions due to long range
interactions is deferred to later chapters. This chapter begins with a brief
summary of the basic physical notions behind short range Rydberg atom-
neutral scattering, followed by an outline of the theoretical connection
between Rydberg atom scattering and free electron scattering.1"5 Commonly
used experimental techniques are described, and a summary of the
experimental results is presented. The closely related line broadening and
shift measurements are described in the chapter immediately following this
one. 195

196 Rydberg atoms Fig. 11.1 A CO molecule colliding with aNa Rydberg
atom which is composed ofthe Na+ core andthediffuse electron cloud
indicated by the shaded area (from ref.6). A physical picture An illustrative
example isthe collision of a Na Rydberg atom with CO, and inFig. 11.1 we
show a picture of the CO passing through the Na electron cloud.6 There are
three interactions: e~�Na+ e~�CO Na+�CO (11.1) The e~�Na+
interaction leads to the energy levels of the Na atom. As long as the
characteristic e"�CO and Na+�CO interaction lengths are small compared
to the size of the Rydberg atom, we can describe collision processes as being
due to the sum of the e~�CO interaction and the Na+�CO interaction,
ignoring any correlation between the two. This approximation isexcellent for
high n states, but is unlikely to be good for n = 5 states. How does the
electron, for example, interact with the CO? It is useful to introduce the
distance pbetween the Rydberg electron and the CO. First, there is the short
range interaction characterized by the elastic scattering length a and a 6{p)
dependence.1 Second, the electron interacts with the electric dipole and
quadrupole moments of the polar CO molecule, yielding p~2 and p~3
dependences for the interactions. Finally, the electron interacts with induced
multipole moments of the CO. The largest of these is the electron-induced
dipole interaction, which scales asp~4. The interactions of the Na+ withthe
CO are analogous except for the short range scattering length interaction. All
the above mentioned interactions are of short range, compared to the size of
an = 20Rydberg atom, so for them to play an important role, the CO must
come near the electron or the Na+. Since the Na+ is localized at a point,
while the electron can be found anywhere in the electron cloud of Fig. 11.1,
the e~�COinteraction is usually, but not always, the dominant interaction.
Above we have considered collisions with CO. If we now consider
collisions with N2, for example, the electron-dipole interaction is absent,
and the longest

A physicalpicture 197 range e~�N2 interaction is the electron-quadrupole

interaction. Since this interaction isof shorter range than the electron-dipole
interaction in CO, it is not unreasonable to expect to see a difference
between Rydberg atom collisions with CO and N2. If we substitute a rare gas
atom, or probably any atom, for the CO, both the electron-dipole and
electron-quadrupole interactions are absent, leaving only the polarization
interaction and the delta function interaction. If we assume that the electron-
perturber interaction is dominant, what can we expect to observe in
collisions between Rydberg atoms and different collision partners? A ground
state rare gas atom has no energetically accessible states, and as a result
collisions between the electron and the rare gas atom are elastic collisions,
involving only the exchange of translational energy. Since an electron
scattering from an atom isroughly comparable to aping pong ball scattering
from a bowling ball, very little energy is exchanged. Specifically, the typical
energy exchanged, AW, is given by \W=4kTv/V (11.2) where k is Boltzman's
constant, T is the temperature, v is the velocity of the Rydberg electron, and
V is the relative velocity of the rare gas atom and the Rydberg atom. Using v
= 1/n, for collisions between n = 20 Rydberg atoms and He at 300 K the
expression of Eq. (11.2) gives AW = 3 cm"1, which is small compared to kT
= 200cm"1. We may thus expect that collision processes inwhich the Rydberg
atom changes energy by more than AW = A;77100 to be unlikely. While this
notion is a good general rule, it is important to bear in mind that in obtaining
this value of AWwe have used a typical value of v ~~ 1/n.However, the
Rydberg electron's velocity is larger near the ionic core, so collision
processes requiring a larger change in the energy of the Rydberg state are
possible, albeit with smaller cross sections. In collisions with molecules the
fact that there are energetically accessible vibrational and rotational states
alters the picture significantly. An electron scattering from a molecule can
induce rotational transitions in the molecule by means of dipole or
quadrupole transitions in the molecule. The dipole transitions are of course
stronger, but only occur in polar molecules, such as CO or NH3. In such
electron induced transitions the energy of the molecule must change by the
amount separating the initial and final rotational states, and as a result the
Rydberg electron's energy must change by the same amount. In other words
this process is necessarily resonant, although the resonant behavior maynot
be readily apparent if there are many rotational states populated. Vibrational
excitation is also possible, and molecular vibrational transitions, which are
easily induced by electron impact, play a role in the depopulation of low
lying Rydberg states. In sum, the presence of the rotational and vibrational
states of molecules allows inelastic collisions of the Rydberg electrons with
the molecule, and this new channel vastly increases the chance of
substantially changing the energy of a Rydberg state in a collision.

198 Rydberg atoms Theory As we have already noted, many Rydberg atom
collision processes are due to the interaction of primarily the Rydberg
electron with the perturber, and much theoretical work, beginning with that of
Fermi,1 has been devoted to connecting electron scattering to Rydberg atom
scattering. For example, detailed reviews have been given by Omont,2
Matsuzawa,3 Hickman etal.4 and Flannery.5 While it is not possible to give
an exhaustive treatment of the theory here, it is useful to present a brief
outline which shows the connection between electron scattering and Rydberg
atom scattering. We are interested in collisions for which the range of the
interaction between the Rydberg electron and the perturber is small
compared to n2a0, the size of the Rydberg atom, for we are assuming that the
Rydberg electron and ion core scatter independently from the perturber.
Furthermore we shall in many cases ignore the interaction of the Rydberg ion
core with the perturber. For low energy electron scattering from rare gas
atoms the scattering length gives a reasonable estimate of the range of the
electron-rare gas interaction. Since typical scattering lengths are ~3ao>7 it is
evident that for any Rydberg state of n > 2 this criterion ismet. On the other
hand, electron-polar molecule scattering is characterized by cross sections of
103 A2,8 so interaction lengths of ~30�0 are not uncommon. Thus for polar
molecules only for n > 5 can we ignore the interaction of the perturber with
the Rydberg ion core. Consider very fast collisions, in which the collision
velocity V between the Rydberg atom and the perturber is much larger than
the Rydberg atom's orbital velocity, v, explicitly, V � v. For collisions with
1 keV H atoms this criterion is met for n � 10. In such a collision the orbital
velocity of the Rydberg electron is negligible compared to V. In effect,
scattering from a Rydberg atom with velocity V is identical to the scattering
from a nearly static electron cloud plus the scattering from the Rydberg ion
core. Explicitly, for fast collisions,3'910 �Ryd-perturber = �e~-perturber
+ ^ion-perturber � (H-^) If we focus on the other extreme, collisions at
thermal energies, the relative velocity of the colliding atoms is ~10~3, and
the typical velocity of the Rydberg electron is v = IIn. Therefore for n �
1000 the Rydberg electron's velocity substantially exceeds the velocity of the
perturber, and a useful way of approximating the collisions isaselectron
scattering from staticperturbers. This approach is valid when both the
requirements of the range being smaller than the orbital radius and v � V are
met. Even for the relatively restrictive case of polar molecules they are met
for 10 < n < 1000. Consider a perturber which collides with a Rydberg atom
as shown in Fig. 11.2(a). Before the collision the perturber is in state (5 and
has momentum K. After the collision the perturber is in state /?' and has
momentum K'. The

Theory 199 (a) (b) Fig. 11.2 (a) Schematic diagram of aperturber p initially
in state /? and having momentum K scattering from a Rydberg atom to
produce final state /3f and momentum K'. (b) Magnified view of the Rydberg
electron colliding with the perturber. The Rydberg electron is initially in
state a and has momentum k. After the collision these are a' and k\ perturber
moves slowly through the Rydberg atom until it is hit by the more rapidly
moving Rydberg electron, which causes an abrupt change in the momentum
and state of the perturber to K' and /?' respectively. In Fig. 11.2(b) we show
the trajectory of the Rydberg electron just before and after hitting the
perturber. Prior to the collision the Rydberg atom is in state a,

200 Rydberg atoms and the electron has momentum k; after the collision these
are a' and k', respectively. In Fig. 11.2(a) wehave drawn the scattering angle
6 as afairly large angle,for clarity, but, sincethe Rydberg electron cannot
exchange much momentum with the perturber, 6 is in reality quite small.
Accordingly, we begin by treating the problem using the Born approximation.
If we define the momentum transfer Q = K � K', the Born scattering
amplitude for the collision of Fig. 11.2 is given by3'11 /(a,^K^a',y3',K') =^|ei
QR ^w (r)f/08^',r,R)^m (r)d3 r d3R, (11.4) where Uis the interaction
potential for initial andfinalperturber states/? and/?', r and R are the positions
of the Rydberg electron and the perturber relative to the Rydberg ion core,
and ju is the reduced mass of the Rydberg atom and the perturber.
InwritingEq. (11.4)we have chosen theusual sphericaln �mstates for the a
Rydberg states. Throughout this sectionwe shall assumethat the a states are
spherical states,but thisisnot the onlypossible choice.Forexample, for
hydrogen in afieldthe parabolic states would be abetter choice.12Sincewe
are considering the electron-perturber interaction, in general U(J3,fi',r, R)
must be afunction of r � R, soweintroduce p = r - R and rewrite Eq. (11.4)
substituting p 4-r for R. This procedure yields /(a,/3,K^ a,/3,K') = ^ |
eiQ>*.rm'(r)f/O8,y3',p) e-iQ"Vn�m (r) d3rd3p(11.5) Examining Fig.
11.2(b), we can see that by conservation of momentum K + k = K'+k'. (11.6)
Thus Q = K � K' implies that Q = k' � k. In other words, Q is the
momentum transfer to the electron. If we assume that there isno particular
orientation of the perturber during the collision we canreplace U(/?,/?', p)
bythe isotropic potential U(J3, /?', p). With this approximation in Eq. (11.5)
we recognize the Born approximation to the electron scattering amplitude / 0
M ' Q ) ( - i Q < W / J ' ) d 3 (11.7) m and the form factor F(a,a',Q) =
<t/VrJeiQr|Wm> = j VnVm' (r) eiQ' *y>ntm to d3r (11.8) InEq. (11.7) mis the
massof the electron, i.e. 1 inatomicunits. UsingEqs. (11.7) and (11.8) we may
rewrite Eq. (11.5) as m
Theory 201 For polar molecule perturbers the Born electron scattering
amplitude is quite accurate and Eq. (11.9) is immediately useful. As an
example, the squared Born scattering amplitude for / �> / � 1 rotational
deexcitation of a polar diatomic molecule is given by3 " C D i r ' r ' x " 3
Wo/ (2/+l)02 where0 = /, 0' = /-I, Dis the permanent dipolemoment of the
molecule, and eis the electron's charge. TheBorn amplitudes for other polar
moleculeshavesimilar forms and are given by Matsuzawa.3 The presence of
1/g2 in Eq. (11.10) reflects the long range of the electron-dipole interaction.
In contrast, the Born amplitude for electron scattering is not always accurate,
especially for lowenergy scattering from rare gases,inwhich the scattering
angles are large. A natural approach is to allow the replacement of Born
amplitude of Eq. (11.7) by a more accurate electron scattering amplitude/e
(00'Q).13 In other words, we express the scattering amplitude of Eq. (11.5)
as Including the electron mass explicitly in Eq. (11.7) yields the ratio ju/m in
Eq. (nil). The electron-rare gas scattering amplitude is given by6 (11.12)
where rj is the s wave phase shift, which is related to the scattering length
aby6 tanrj(k) = -ak. (11.13) If we assume that rj is small, 4(0, 0, Q) = �a,
and the total electron-perturber elastic scattering cross section is given by
oe(Q) = In[" \fe(fi, 0, Q)|2sin 6dO (11.14) Jo = Ana2. For arare gas atom it is
useful to write out the coupling between states a and a' in a fashion analogous
to the scattering amplitude of Eq. (11.4). Explicitly, (a' \U(P,P, r, R)|a) = L j
W (r)�7(0,0, r, R)ipn�m (r) d3r. (11.15) If we substitute the Fourier
transforms of the spatial wavefunctions and use the fact that U(j3,0, r, R) =
�7(0,0, p), we can rewrite Eq. (11.15) as <a'|�7|a) = fe-
ik'rG2'(k')�/(0,0,p)Ga(k) eikr d3rd3�d3�' (11.16) substituting R + pfor r
yields <a'|�7|a>= [e"i(k'-k)RG*,(k')Ga(k)ei(k'"k)p�/(0,0,p)d3p d3kd3kf.
(11.17)

202 Rydberg atoms The integral over p is the Born amplitude for electron
scattering multiplied by -2JZ. Using/e(/?,�, Q) = -0, replacing theintegral
over p of Eq. (11.17) by 2jza, and integrating over k and k' yields (a'\U\a) =
<VV*'m'ltfIVWm>= 2jta V^'m'(R)VWR) (11.18) Inspecting Eq. (11.18),
wecan seethat we canequivalently express U(j3,/3, r, R) as �,r9 R) = 2jtad(r
- R). (11.19) The form factor of Eq. (11.8) is an integral over three factors,
two wavefunctions and an oscillatory plane wave term, elQ r . Due to the
plane wave term the magnitude ofthe form factor decreases with increasing
Q,and it may be evaluated using either the methods of Gounand and
Petitjean14 or Cheng and van Regemorter.15 In analternative approach16
used by Hickman,17 the form factor is replaced by a Bessel function
expansion. To compute the total cross section we integrate the squared
scattering amplitude over the scattering angle 0. Explicitly, (11.20) Using the
fact that Q2 = K2 + K'2 - 2KK' cos 6, (H-21) we can replace the integral
over the scattering angle 6 by an integral over Q, Qm
�'\F(a,a',Q)\2feWQ)\2QdQ, (11.22) i.e.3-11 where <2min = |K� K'
|andQmax = |K+ K' |. Since almost allthe scattering isnear the forward
direction Qmax can be replaced by �� with minimal error. Energy
conservation determines <2min- Explicitly,311 ^ + Wa + Wp =^ + W� + Wl,
(11.23) where Wa, Wa>, Wpy andWp> are the internal energies ofthe
Rydberg atom andthe perturber before and after the collision. Solving Eq.
(11.23) for |K � K'| in the approximation K ~ K' yields Qmin = |K - K'| =�|
(Wa - Wa.) - {Wp - Wfi.)\. A (11.24) There are two important points tonote
regarding Qmin, and Eq. (11.24). First, the right hand side contains the
difference between the internal energy lost by the Rydberg atom and that
gained by the perturber. If this difference iszero, so that the collision
isresonant and no energy goes into translation, Qmin = 0.Second, for an
appreciable cross section <2min should besmall, since the form factor
decreases

Theory 203 with Q, andinthecase ofpolar molecule perturbers the scattering

amplitude,/eB, does aswell. Consider two examples. For a rare gas perturber
there canbe no change in the state of the perturber, so /? = /?' and <2min is
determined entirely by the energy difference, Wa � Wa>, between the initial
andfinalRydberg states. On the other hand, ifthe perturber is adiatomic
molecule which isrotationally deexcited in the collision so that Wp < Wp and
the Rydberg atom's energy gain equals the molecule's energy loss, i.e. Wa> -
Wa = Wp - Wp>, Qmin =0 in spite ofthe fact that Wa*Wa.. The cross section
of Eq. (11.22) isfor a single initial andfinalstate. Assuming that we have no
control over the direction of the collision velocities, tocompute a cross
section wemust, at a minimum, average over m of the initial state and sum
over m' of the final state. This procedure leads to i �
^\F(ntm,n't'm',Q)\2\fe(P,P',Q)\2QdQIn KK> JQmin^ ^ w (11.25) Eq. (11.25)
is in somewaysthe minimally useful crosssection, from one nt state to another
n't' state. If we are interested in the total depopulation of the nt state we must
sum over the possible n't' final states. The summation over n't' includes,
implicitly, an integration over the continuum, although including the
continuum is usually unnecessary. In the continuum \F(ntm, n't'm' ,Q)\2n'~3 is
replaced by d\F(ntm,W't'm',Q)|2/dW, where d\F(ntm,W't'm',Q)\2 _. \F(ntm,
n't'm',Q)\2 H 1 , . = lim = � (11.26) dW �,_>� n'5 The squared form factor
is similar to the oscillator strength in that it passes smoothly across the
ionization limit. If we are interested in ionization we simply sum over t' and
integrate over the continuum using Eq. (11.26). The approach outlined above
has been used extensively by Matsuzawa3 to calculate cross sections for
excitation and ionization of Rydberg atoms by polar molecules and by
Hickman4 to calculate cross sections for state changing by rare gas atoms
colliding with Rydberg atoms. Another approach to theproblem of rare gas
scattering is to replace the spatial wavefunctions of Eq. (11.4) with their
Fourier transforms, the momentum space wavefunctions. These
wavefunctions represent the velocity distributions of the electron inthe
Rydberg states. Proceeding along these lines,werewrite Eq. (11.4) as11 x
ei(k " k '} *r�/03, p\ r,R)Ga(k) d3k d3k' d3r d3R (11.27)

204 Rydberg atoms Replacing r by p + R, assuming U (/?,ft',r, R) = U(J3,

/?',p), and integrating over R yields x e"ik'' p �/08,0', p) eik"p d3p d3fcd3fc'
(11.28) In addition to showing the fact that the momentum transferred to the
collision partner comes from the electron, this expression also contains,inthe
integral over p, the Born scattering amplitude, given by Eq. (11.7). If we
again replace it by the correct amplitude/e(/J, >8', k' � k)12 and integrate
over k' wefind = ^ J G*,(k + Q)Ga(k)�(/?,/?', Q) A3 This expression for the
scattering amplitude, given in terms of the momentum, or velocity,
distribution of the Rydberg electrons, is usually termed the impulse
approximation. Examination of Eq. (11.29) shows the similarity of the
integral over the momentum to the form factor of Eq. (11.8). It is useful to
consider rare gasscattering, for which a = a' and/? = /?'. Using the optical
theorem we can relate the imaginary part of the forward scattering amplitude
to the total cross section.18 Explicitly, we can apply the optical theorem to
the scattering amplitudes on both the left hand and right hand sides of Eq.
(11.29). This procedure yields ^(aJ,K)\\Ga(k)\^oe7(P,k)d3k. (11.30) An m)
An Using K/JU = V and klm = v, Eq. (11.30) reduces to V Oj{J3 9K) = J
|Ga(k)|2vaeT (0, k) d3k. (11.31) The physical significance of this expression
is that the rate constant for elastic perturber-Rydberg atom collisions is the
same as the rate constant for elastic perturber-electron collisions averaged
over the Rydberg electron's velocity distribution, in spite of the fact that V �
v. Eq. (11.30) isstrictly applicable only to elastic collisions, inwhich a'' = a,
and is thus of limited utility. However, it isphysically appealing to assume
that the cross section a(a, ar, /J,ft')for an inelastic process a' ^ a and /?' ^ftcan
be written as the integral of the electron scattering cross section ae(/J, /?', q)
over the velocity distribution of the Rydberg electron in the initial state a.
Making this notion explicit, we write19 Vo1(a9a',l39pt,)= f |Ga(k)|2va0M',
k)d3k, (11.32) where kmin accounts for the fact that not all momenta present
in the initial state wavefunction contribute to the scattering. Explicitly,

Experimental methods 205 i�Wfi'>Wfi (11.33) 0 i�W'<W As is the case

for Eq. (11.31), Eq. (11.32) states explicitly that the rate constant for
Rydberg atom scattering is the same as the rate constant for free electron
scattering averaged over the Rydberg electron velocity distribution. It is
instructive to consider applying Eq. (11.32) to processes in which the
electron-perturber scattering is short range and longrange. When the
interaction is short range, as in elastic electron-rare gas scattering, the
electron-rare gas scattering cross section OGT, has no velocity dependence
and is simply given by And?. In this case the cross section for Rydberg atom
scattering given by Eq. (11.32) can be written as ^ (11.34) where the
expectation value of the electron velocity, (v) is given by (v)=f|Ga(k)|2vd3/:.
(11.35) The cross section of Eq. (11.34) scales as IIn. In contrast, for a long
range interaction, such aselectron-polar molecule scattering, the crosssection
<7erhasa 1/v dependence and the electron scattering rate constant vaeris
independent of v, leading to an n independent cross section OJ(J3, K) and
rate constant VoT(fi, K). The notion that the Rydberg atom scattering rate
constant is equal to the electron scattering rate constant averaged over the
Rydberg state velocity distribution has proven to be very useful in, for
example, treating collisions with attaching targets.20'21 It also forms the
basis for the method of de Prunele and Pascale for treating �mixing
collisions.22 Experimental methods There are three methods which have
been used to study collisions of Rydberg atoms with neutrals. They are direct
measurement of collisionally induced population changes, line shift and
broadening measurements, and photon echo measurements.23 In this chapter
we describe the first of these. The last two are described in the chapter
immediately following. The most commonly used method isthe direct
measurement of a decay rate by pulsed excitation and time resolved
detection. The most straightforward example of this technique is laser
induced fluorescence applied to alkali Rydberg atoms. Alkali atoms are
typically contained in a glass cell, which also holds a known pressure of
perturber gas. The alkali atoms are excited to the Rydberg state at time t
=0and the time resolvedfluorescencefrom the Rydberg atoms isdetected

206 Rydberg atoms at t > 0. Assuming that the alkali ground state atoms are at
alowenough pressure that they do not depopulate the Rydberg state, the
intensity of the fluorescence with no perturbing gas is given by /=/o e-V
(11.36) where y0 is the inverse of the radiative lifetime at the cell
temperature, which usually includes a contribution due to black body
radiation, as discussed in Chapter 5. When a perturbing gas is added
thefluorescenceintensity is givenby I=Ioe-(yo+rc)\ (11.37) where yc is the
rate at which collisions remove atoms from the state excited by the laser. The
collision rate dependsonthedensity,orpressure, oftheperturbing gas.
Explicitly, yc = nk =noV, (11.38) where n isthe number density of the
perturbing gas, a and k are the cross section and rate constant for
depopulation, and V is the average relative velocity of the Rydberg atom and
the perturber. In a cell of temperature T, V = y/8kT/jrju, k isBoltzman's
constant and// isthe reduced mass of the collision partners. Since the number
density isproportional to the pressure, the collisional contribution to the
decay rate increases linearly with pressure, and measuring the decay rate asa
function of pressure, as shown in Fig. 11.3, yields the cross section.24 Such a
measurement really yields the rate constant k, but since it is more physically
appealing to compare a crosssection to the sizeof an atom we shallmost often
use cross sections derived using Eq. (11.38). However it should be borne in
mind that the cross section so obtained is an average over thermal velocities.
The approach described above is the simplest variant of time resolved
fluorescence detection for the measurement of collision processes. If we
choose to excite state A and detect thefluorescencefrom a second state B, the
fluorescence intensity has the form / = /0 (e-^-e-V), (11.39) where ys and yf
are the slower and faster of the decay rates of the two statesA and B. Eq.
(11.37) is written assuming that after atoms are removed from the state
excited by the laser, state A, they never return. In contrast, if the atoms
initially excited to state A are collisionally transferred to a longer lived state
R and do return toA, thefluorescencefrom stateA exhibits atwoexponential
decaywhich is of the form25 / = /0(a e"^' + p e-y**), (11.40) where a and/3
depend on the decay rates yA andyR. Whenever yA andyR are at all close to
each other it is difficult to extract good values from even slightly noisy data.

Experimental methods 207 0.2 0.4 0.6 0.8 1.0 1.2 Pressure (torr) Fig. 11.3
Decay rate of the Na 8s state vs N2 pressure, obtained by observing the time
resolved 8s-3pfluorescencesubsequent to pulsed laser excitation (from ref.
24). Fluorescence detection is simple, and if the detected fluorescence is
well resolved in wavelength there isno doubt about which state isbeing
observed. On the other hand fluorescence detection does have its limitations.
As n increases, the number of Rydberg atoms excited decreases as n~3 as
does the spacing of the levels, so that increasingly higher spectral resolution
is required. Although the time integrated intensity only declines as n~3, the
intensity of the fluorescence signals decreases as n~6, and background noise
becomes more of aproblem with increasing n. In addition, it is rarely
possible to collect even 10% of the fluorescence emitted at any given
wavelength. Finally,fluorescenceradiated on unobserved transitions goes
undetected, so the branching ratio for the transition monitored must be
favorable. In light of the inherent difficulties of fluorescence detection at high
n, it is not surprising thatfluorescencedetection has only been used torn <
22.26 The other commonly used technique is selectivefieldionization. The
atomscan be in abeam or in acellof the type shown inFig. 11.4, in whichthe
detector is in a separately pumped region connected to the interaction region,
which may have a high pressure, ~10~3torr, of added gas.27The
basicprinciple isto use the known

208 Rydberg atoms Ionic signal To pump ! h i � Electron multiplier

Hole1mm diameter � _ _ . _ _ _ /"" ~ j \ Laser interaction zone
Experimental cell . , . _ . Gas in Electric field plates 1 � HV Fig. 11.4 Two
chambered cell allowing the use of field ionization to study collision
processes.Thelower chamber contains a relatively high pressure
andtheupperchamber, containing theelectron multiplier, isat a lowerpressure
(from ref. 27). values of thefieldsat which states ionize to connect, using the
principles outlined in Chapter 7, a signal at a givenionizingfieldto aspecific
state. If asingle stateA is excited, andthe field ionization ramp is applied at a
variable time t later, the Rydberg states present attime tmay be determined by
simply noting the times, and thus the fields, at which signals are observed.
Repeating this procedureon many shots of the laser for different values of t
allows one to build up a record of the time resolved populations of the
Rydberg states. The attractions of field ionization are two. First, it is
efficient. All the Rydberg atoms are ionized, and all the resulting ions or
electrons can be detected with high efficiency. Second, allthe atoms present
are ionized at once, so the signals have an n~3 scaling, and background noise
isusually not aproblem. On the other hand, the successful use of field
ionization is always dependent upon correctly connecting the ionization field
to a state, as described in Chapter 7. Collisional angular momentum mixing
One of the more well studied collision processes involving Rydberg atoms is
collisional angular momentum mixing, or �mixing, the collisional transfer of
population among the nearly degenerate �states of the same n.2S The
process has

Collisional angular momentum mixing 209 100, 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4
1.6 1.8 2.0 Fig. 11.5 Semilogarithmic plot of theinitialportion ofobserved
decayoftheNa lOd state in the presence of 0.027 Torr Ne (�). This plot
shows the entire fast decay but only the beginning of the slow decay. The
solid line is the computer fit of two exponentials to the data. The
decaytimesofthefast andslow component from thecomputerfitare0.19and3.9
jus, respectively (from ref. 25). been studied in alkali atoms and Xe, and it is
limited to � states with quantum defects near zero. For example, rapid �
mixing has been observed from the Li nd,29 Na nd,25'30'31 Rb ni,32 and Xe
ni33'34 states to the higher �states of the same n. Most of these experiments
have been done by laser induced fluorescence techniques. Typical of the
observations is the decay of the Na 10d-3p fluorescence in the presence of
Ne, shown in Fig 11.5.25 The decay clearly contains more than one
exponential, and, infact, isfitwellbyatwoexponential decay.Thefast initial
component is the decay of the initially populated lOd state, and the slow
component is the apparently pressure independent decayofthemixture of n =
10, � > 2 states. The identification of this process as � mixing was based
on two observations. First, the same rare gas pressure had no observable
effect ontheNa ns states which have a quantum defect of 1.35. Second,
measuring the pressure independent slow decay rate, such as the one shown
in Fig. 11.5 gives a lifetime rnocn4'5^6\ Thisn dependence is in agreement
with the expected n4'5 scalingof the average lifetime of all �,m states of the
same n.35'36 When the missing ns and np states aswell as the difference
between the radiative lifetimes of the Na and H nd states are taken into
account the observed values of rn are in excellent agreement with the
calculated values.25 In retrospect, the agreement is probably somewhat
fortuitous. Thermal radiation shortens the lifetime of the mixture of �> 2
states populated by collisions of rare gas atoms with the initially excited nd
states. However, the most likely transitions from an ni state are to w�l, �
� 1states. Since the Na nd-3p fluorescence was observed through a filter
which would

210 Rydberg atoms 16 20 Fig. 11.6 Observed (O) (ref. 25) and calculated
(�) (ref. 37) cross sections for the collisional �mixingoftheNandstatesby
He, Ne, andAr. The calculated crosssections are based on the two state
model of Olson (from ref. 37). transmit the (n�l)d-3p fluorescence as well,
the effect of black body transitions between ( states of similar n was not
noticeable. The � mixing cross sections are obtained by measuring the
decay rate of the initial fast decay asafunction of perturbing gas pressure. In
Fig. 11.6 we showthe experimental Na nd �mixing cross sections for the
rare gases He, Ne, and Ar,25 and the cross sections calculated by Olson,37
using atwo state model in which the coupling arisesfrom the scattering length
and the polarizability of the rare gas. As shown, the agreement between the
calculated and observed cross sections is excellent. The calculated cross
sections of Fig. 11.6 are representative of calculations of �mixing carried
out by de Prunele and Pascale,22 Gersten,38 Derouard and Lombardi,39 and
Hickman40'41 as well. As shown by Fig. 11.6 the cross sections rise
approximately as the geometric cross section, n4, at low n, reach a peak, and
then decline at high n. It is apparent that the location of the peak and how high
it becomes depends on the rare gas, and that the peak � mixing cross section
increases from Ne to Ar. Not surprisingly, the cross sections for ( mixing of
the Na nd states by Xe measured by Kachru etal.30 rise to an even higher
peak value of 4 X 10~12cm2 at n = 18. In classical terms �mixingcanbe
thought ofin a simple way. When anAr atom, for example, enters the Rydberg
electron's orbit the electron can scatter elastically from the Ar. The
reorientation of the orbit corresponds to a new �. If we pursue this notion
alittle further wecan understand qualitatively the dependence
Collisional angular momentum mixing 211 onnand the identity ofthe rare
gas.The couplingbetween the �stateswhichleads to �mixing isprovided
predominantly bythe short range interaction between the electron and the rare
gas atom. Thus, to a good approximation, the coupling is given by Eq.
(11.18) i.e. <a'|t/|a) = 2jrfl/^(r)<J(r-R)^a(r) d3r, (11.41) where a and a1 are
the initial and final states, r and R are the locations of the electron and the Ar
relative to the Na+, and a is the Ar scattering length. In other words, the
strength of the coupling depends on the rare gasthrough its scattering length. If
the Ar atom passes through any part of a low n Rydberg atom the strength of
the coupling is enough that � mixing occurs, and the � mixing cross section
equals the geometric cross section of the Rydberg atom. On the other hand, as
n increases the coupling decreases because the wave function becomes more
dilute, and the mere presence of the Ar atom inside the Rydberg electron's
orbit does not ensure an � changing collision. In this regime the i mixing
cross section falls farther and farther below the geometric cross section and
begins to decline with increasing n. The n at which the maximum cross
section occurs increases with the absolute value of the scattering length of the
rare gas. For He, Ne, Ar, and Xe the scattering lengths are 1.19tfo> 0.24tf0,
�1.7(k0, and �6.5a 0. 6 As expected, the maximum cross section for Ne
occurs at the lowest n and the maximum crosssection for Xe occurs at the
highest n. At high n the �mixingcross section isproportional to the product of
the probability offindingthe electron at anypoint in itsorbit, ~n~~6,and the
length of the path of the Ar atom through the Rydberg atom, ~/i2, yielding a
cross section which decreases as n ~4. In addition to the measurements of
total ( mixing cross sections, it has been possible to analyze the final states
subsequent to � mixing. As suggested by the simple classicalpicture and the
coupling matrix element of Eq. (11.41), there is no reason to expect any
�selection rule for �mixing.Usingfieldionization Gallagher
etal.42observed that the �mixingof theNa 15dstatebyArresultsinan
apparently even mixture of n = 15 �>2 states ofm<2. The distribution of \m\
> 2stateswas not observed. Since theNand � niintervalis roughlyfivetimes
greater than the ni � ng interval thepossibility of aslow,rate limiting 15d-
n>15ftransfer followed by a rapid 15f ��15� > 3 transfer cannot be
excluded on the basis of the above experiment alone. However, using
resonant microwave fields to form 50-50 mixtures of Na nd and ni states
Gallagher et al. observed that the ( mixing rateof the nd � nimixturewas
nearly unchanged from that ofthend state,indicating that the �mixing cross
section of the Na ni states isvery nearly the same asthat of the Na nd states, in
spite of the small ni-ng energy separation.42 Later measurements by Kachru
et al.30 and Slusher et al.43 have confirmed the lack of a A� selection rule.
The experimental evidence also seems to indicate that there is no strong Am
selection rule, although there maybe aslight propensity for Am to be small.
First, the fact that the observed pressure independent lifetime of the mixture
of �states

212 Rydberg atoms scales as n45 implies that all \m\ values are eventually
populated. However, several collisions could be required to reach an even
distribution over \m\. More direct evidence that there is no Am selection rule
is provided by the work of Slusher et al.44 First, they observed that the
diabatic field ionization signal subsequent to �mixing of the Xe nf states
matched the distribution expected for all \m\ states, not simply low \m\ states.
Second, in another experiment they allowed the Xe 31f state to be � mixed
by CO2 molecules, producing distinct adiabatic |m|<3 and diabatic
|m|>3fieldionization signals. They then selectively field ionized the low
\m\atoms responsible for the adiabaticfieldionization signal and observed the
rate at which the low \m\states were repopulated from the high \m\ states, i.e.
the rate of Am collisions. They found the rate constant kAm = 2 x 10~7cm3/s,
virtually the same rate as kM = 1.5 x 10~7 cm3/s. In the �mixing of the Na
18d3/2state byXe Kachru etal.30 found that when half the atoms were
removed from the initiallypopulated 18d state roughly 30% of the atoms
removed were in states which ionized at fields lower than the ionization
fieldof the 18d3/2 state. These are states of \m\ ^ 2. The remaining 70% of
the atoms ionized at fields in excess of the ionization field of the 18d3/2
state. These atoms must be in states of \m\ > 3. Counting the \m\states of n =
18 and � > 2 reveals that 23% of the � > 2 states have\m\ < 2, and 77%
have \m\ > 2, so the observed distribution, with 30% of the atoms having \m\
^ 2, shows at most a slight propensity for small Ambut no evidence for any
strong Amselection rule. The effects of electric fields on � mixing The
�mixing of both the Na nd and Xenf states hasbeen studied in electric fields.
Both these states are the lowest states of nearly degenerate high �states, and
both are thus adiabatically connected to the lowest member of the Stark
manifold. The experiments are done by exciting the Na nd or Xe nf states in
zero field then applying a field which rises slowly to a chosen value and then
remains constant while collisions are allowed to occur. In this way only the
lowest member of the Stark manifold is excited. After the l-10jus period
during which collisions occur, the time resolved field ionization signal from
this initially populated level is observed as a function of perturbing gas
pressure to determine the total cross section. As the field is raised from zero
to approximately l/3n5 the cross section drops steadily. Kachru et al.30show
a factor of 2decline for afieldof l/30n5, and Slusher et al43 and Chapelet et
al31 show that the cross section continues to decrease with increasing field,
the decrease reaching a factor of 4 for a fieldof l/3n5. Equally as interesting
as the size of the total cross section is the distribution of the final states
subsequent to � mixing. Examining the adiabatic field ionization signalsof
Kachru etal.,30it appears that onlythe lowest Stark states nearest to the

�mixing by molecules 213 one initially populated are populated by (

mixing. In the data of Slusher et al.43 this point is very clear. As shown by
Fig. 11.7, when the Xe 31f state isallowed to remain in zero field the
diabatic field ionization signal of the products shown in Fig. 11.7(a) exhibits
the form characteristic of having contributions of all�and m states. Recall
from Fig. 6.10 that the lowestfieldportion of the signal comes from low
\m\redmost Stark states. As shown by Fig. 11.7raising thefieldremoves the
higher field portion of the field ionization signal, indicating that the presence
of thefieldrestricts thefinalstates to the low lying Stark states of lowm
adjacent to the one initially populated. There are two effects which come into
play in an electric field. First, the Stark states are polarized along
thefieldaxis, the red and blue Stark states having their wavefunctions
primarily in the downfield or upfield directions. The consequence of this
spatial character is that the matrix elements of the short range e~-Xe
interaction given by Eq. (11.41) are only appreciable between relatively
similar Stark states. This effect alone would certainly restrict the number of
final states with appreciable populations, as observed. Hickman's
quantitative theoretical description of collisions in electric fields, using
parabolic Stark states, also indicates that only relatively similar Stark states
are populated.12 The final state restriction might ormight not alter
thetotalcrosssection depending upon whether or not the spatial overlap with
nearby Stark states wasimproved enough to offset the smaller number
offinalstates. However, this effect cannot explain the steady decrease in cross
section with increasing field, for the Stark wavefunctions are
fieldindependent. This realization brings us to the second important effect of
the field, the limitation of kinetic energy transfer. According to our previous
argument regarding kinetic energy transfer between an electron and aHe
atom, we expect typical energy transfers of <1 cm"1. Since the energy
between adjacent Stark states is 3nE, thisconstraint reduces the number
offinalstates as thefieldisincreased, and it is for this reason that the cross
section decreases monotonically with field, at least for E < l/3rc5.43 �
mixing by molecules As discussed above, molecular collision partners have
energetically accessible degrees of freedom and longer range interactions.
What effect do these have on � mixing? While not as much effort has gone
into the measurement of �mixing by molecules, a reasonable range of
measurements has been carried out. For nonpolar molecules and only slightly
polar molecules such as CO, the behavior is qualitatively the same as a rare
gas atom's. The cross sections for �mixing of the Na nd states by CO and N2
have been measured and are nearly identical to the cross sections for Ar.45
The vibrational and rotational degrees of freedom play no role, because the
Na Rydberg �states being mixed are degenerate on the scale of

214 Rydberg atoms i 1 1 1 1 r \ 0 Vcm"1 6 Vcm-1 74 Vcm-1 91 Vcm"1 n=3l

n=30600 800 1000 1200 1400 IONIZING FIELD (Vcm"1) Fig.
11.7CoUisionally induced diabatic ionization features observed following
collisions of laser-excited Xe 3If atoms with Xe target gas at a pressure of
10"5 Torr for 8 jus in the presence of several applied static fields. The
arrows indicate the ranges of ionizing field strengthsoverwhichn = 30 and31
statesareexpected toionizediabatically (from ref. 43).

Fine structure changing collisions 215 molecular rotational and vibrational

frequencies. Since no rotational or vibrational transitions of the molecule are
induced, the long range electron-dipole and electron-quadrupole interactions
are unimportant. Highlypolar molecules, such as HC146and NH334havevery
large �mixingcross sections, asshown byStebbings etal.It is not clear
whether thepolar molecules interact with only the electron ofthe Rydberg
atom orwith the atom as awhole. Fine structure changing collisions In the
heavier alkali atoms, Rb and Cs, thefinestructure intervals of the nd states
can be resolved well enough that population transfers between these states
can be observed. Typically these measurements have been done bytime and
wavelength resolved laser induced fluorescence, for example populating the
nd3/2 state and observing thefluorescencefrom either the nd3/2 or nd5/2
state. Since the nd states of Rb and Cs are energetically well removed from
the high �states the depopulation of the initially populated ndj state is
usually due to fine structure changing collisions, which have large cross
sections. For example, as shown by Deech et al.47 and Pendrill48,
thefinestructure changing crosssection forthe Csndstates of n =9-15 by ground
state Cs atoms increases as n*4, thegeometric cross section. This is the same
dependence as the �mixing cross sections at low n. The n*4 dependence
ofthe Csndfinestructure changing collisions wasconfirmed by Tarn et al49
using a rather different technique. They used continuous wave dyelaser
excitation toexcite Cs6p1/2atoms tothe nd3/2 level and measured the ratio
ofthe fluorescenceradiated on the nd3/2 �> 6p3/2 and nd5/2 �> 6p3/2
transitions. Aside from the 6dstate, which has an anomalously large
crosssection, the crosssections far the process nd3/2�> nd5/2 follows an
n*4 dependence, rising from 3.7(7) x 1(T14 cm2 at 7d to 29.8(60) x 1(T14
cm2 at 10d.49 A set of measurements similar to those of Deech etal.47 was
carried outby Hugon et al. ,50 who measured the Rb nd3/2-> nd5/2 cross
section for n = 9,10 and 11, in collisionswith He. Theyfound that the
crosssection decreased from 5.1(10) x 10"14 cm2 at n = 9 to 2.2(6) x 10"14
cm2 at n = 11, behavior similar to the He-Rb ni �mixing cross section in the
same n range. Furthermore the cross section is smaller than the Rb ni �
mixing cross sections by a factor of 2-3.32 In contrast to the readily observed
fine structure transitions in the Rb and Cs nd states, when Gallagher and
Cooke attempted toobserve them in the Na nd states using rare gas collision
partners they were unable to do so. Specifically, they attempted to observe
the Na nd5/2�� ni7/2 microwave transition when the Na nd3/2 state was
populated and rare gasadded. However theywerenever ableto observe the
transition due to itsbeing obscured by �mixing collisions.51 All of the
above observations are consistent with interpreting fine structure changing
collisions inRydberg atoms as elastic e~-perturber scattering leadingto

216 Rydberg atoms (n + 2)P (n + 2)S (n nD (n + 1)P (n - 2)F,G,H, (n + 1)S

nP (n - 3)F,G,H,... (n - 2)D (n-1)P (n - 4)F,G,H,... Fig. 11.8 Rb energylevel
diagram showingtheproximity ofthe (n+ 3)s statetothe��,� > 3 states and
the relative isolation of the np and �d states (from ref. 50). m but not �
changing, which has been shown to have basically the same cross section as
�mixing.42 Similar n dependences of the cross sections are observed in
both cases. The only real difference is that infinestructure changing collisions
the number ofpossiblefinalstatesisreduced. However,
theoverlapoftheradialwave functions is 100%, or, equivalently, the form
factor is 1. n changing collisions by rare gases We consider first the simplest
case, n changing collisions with rare gas atoms, where by n changing
collisions we mean collisions in which the initial Rydberg state is not
approximately degenerate with other nearby states as is the case in (, mixing
collisions. Since there are no energetically accessible internal states of an
atomic collision partner, it is reasonable to expect the cross sections to be
small since energy must go into, or come from, translational kinetic energy.
Many n changing cross sections have been measured. For example, the
depopulation of theNans statesbyseveralrare gases hasbeen studied over
awide range of n values.52"54 To convey the essential ideas, though, it is
useful to consider a single set of measurements, the depopulation cross
sections of Rb Rydberg states with He. In Fig. 11.8 we show the energy
levels of Rb.50 The s, p, and dstates allhave substantial quantum defects, and
the fstateshave a quantum

n changing collisions by rare gases 217 Fig. 11.9Cross sections for the
depopulation of the Rb nt states byHevs n; ns(O) ( ref. 50,55),rcp(#)( ref.
26),/id(+) (refs. 50,55),andwf(x) (ref. 32). defect of 0.05. We begin with the
Rb ni states, which undergo primarily t mixing collisionspopulating the
degenerate ni >3 states.TheHe �mixingcross sections, shown in Fig. 11.9,
rise to a maximum of 1050A2 at n = II.32 The cross sections for
depopulation of the mixed pair of ndfinestructure levels aremuch smaller,
asshownbyFig. 11.9. Thelown, <15, cross sections, measured by
fluorescence detection, rise to their maximum of 130 A2 at n = 15,50 and the
higher ncrosssections, measured usingfieldionization, decrease from 118 A2
atn = 30to 59A2 at n = 41,55 Although there are no measurements for 15< n
<30, from the data in Fig 11.9 it is hard to imagine that the cross section ever
exceeds 200 A2. The nip state depopulation cross sections were all
measured using fluorescence detection, and asa result the highest nobserved
is n =22.26Asshown byFig 11.9, the crosssectionsriseto aplateau of60 A2,
avaluedistinctly smallerthan thecross sections for any other nt states. Finally
the Rb nscrosssections for 12 < n < 18 were measured bylaser induced
fluorescence and showasharpincreasewithn, to320A atn = 18.49Forn = 32to
n

218 Rydberg atoms 10-13 o o* 1O"M �Na13F I � r Na13D \ * Rb13F N^

� i � . . i t i \ Rb16Si\ F i i i i n . I J Rb15D ^b16P^ I10"15 10"1 1 10 100
AW (cm"1) Fig. 11.10 Log-log plot of the cross sections for quenching of Rb
(x) and Na (O) levels having approximately n* ~ 13by He against the energy
defect AW between the level of interest and its closest state (from ref. 32). =
45 the cross sections were measured using time resolved selective field
ionization,55 and in this n range the cross section decreases from 125 A2 to
82 A2. Inspection of Fig. 11.9 suggests that the 320 A cross section of n =
18may well represent the peak cross section. Itisuseful tomake somegeneral
observations about allthecross sectionsofFig. 11.9.For each n( seriesthe
crosssection increases sharplywithn at lown, reaches a peak at n ~ 10-20,
and then declines. Furthermore, it seemspossible that allthe n� states might
have thesame cross section at high n, although this point is not clear from the
data of Fig 11.9. Finally, inthe region of the maxima in the cross sections
there are pronounced differences in the cross sections of different � states.
The difference in the cross sections is directly related to how energetically
far removed an n( state is from possiblefinalstates. This point is made
explicitly by Fig. 11.10,which is aplot of the depopulation cross sections of
Rb n( statesof effective quantum number n* = 13 byHe.32For comparison the
Na 13d and 13f i mixing cross sections are also shown. The cross sections
are plotted as a function of the energy to the nearest state, AW, and, as shown,
the cross sections are inversely proportional toAWfor large values ofAW,but
independent ofAWfor small values of AW. TheNa ml, Na nfy and Rb ni states
all have similar cross sectionsfollowed insize bythe Rbns stateswhich
areenergetically close tothe (n3) � >3states. Mclntire etal.53present agraph
similar toFig. 11.10 making the same point. In Fig. 11.10 the energy
separation used for the d state, the d-s interval, suggests that the d state
isdepopulated by collisional transitions to primarily thes

n changing collisions by rare gases 219 30 40 50 60 70 PRINCIPAL

QUANTUM NUMBER n Fig. 11.11 Cross sections for state changing in Rb
nt � Xe collisions. Rb ns(O) Rbwp(G|), Rb nd(A) (ref. 56);Rb �s(#),
Rbnd(A) (ref. 55)Born approximation calculations for the Rb ns (�) and nd(
) states (ref. 55), degenerate initial and final state calculation (�) (ref. 12)
(from ref. 56). state. In this case one might expect the cross section to be
much smaller than the depopulation cross section of the s state which can go
to any high � final state. However, it does not seem to be. This observation
and the fact that fine structure changing cross sections are comparable to
�mixing cross sections suggest that the relatively good spatial overlap of
states of similar or identical � more than offsets the availability of a greater
number of states of quite different �. Stated another way these observations
imply a propensity for small A�, in contrast to the observation that �
mixing seems to have no A� selection rule. In Fig. 11.9 there is the
suggestion that at high n all the Rb nt depopulation cross sections coalesce.
Goeller et al.56 have demonstrated this point explicitly using Xe to
depopulate Rb ns, np and nd states of 27 < n < 70. Their cross sections,
shown in Fig. 11.11 show clearly that at high n these three nt series have
depopulation cross sections which converge to the same value. In summary, n
changing collisions are typically collisions which transfer an initially excited
Rydberg atom to the nearest manifold of degenerate high �states,

220 Rydberg atoms perhaps via an intermediate state. These collisions can
be understood as arising predominantly from the short range electron-
perturber interaction leading to elastic e~�He scattering. The Rb�He n
changing cross sections have the same qualitative dependence as the �
mixing cross sections. They increase sharply at low n to a plateau and then
fall as n is further increased. There is only a quantitative difference from
�mixing collisions,more energy must be transferred toor from the Rydberg
electron, requiring that the perturber beclosertotheionic core where the
Rydberg electron's velocity is higher. Stated another way these collisions
rely on the high energy tail of the Rydberg electron's velocity distribution. At
very high n, higher than investigated in this type of experiment, the e~-
perturber interaction should fall to the point where the ion-perturber
scattering is dominant. n changing by alkali atoms While the cross sections
for depopulation of excited alkali Rydberg atoms by rare gas atoms depend
heavily on the quantum defects of the Rydberg states, the same is not true of
the depopulation by ground state alkali atoms. In Figure 11.12we show the
cross sections for the depopulation of the Rb ns,ftp,nd, and ni states in
collisions with ground state Rb,26'32'57 as well as the nd3/2finestructure
mixing cross sections.57 As can be seen in Fig. 11.12, the cross sections for
all of these processes are near the geometric cross section 5jm*4/2. The
similarity of the cross sections is surprising in that the states under
consideration range from the ni states, which are essentially degenerate with
the higher � states, to the nd and np states which are well isolated, as shown
by Fig. 11.8. In the same n range the depopulation cross sections by He are
very different for these n� series, asshown by Fig. 11.9. A point about Fig.
11.12 which is worth noting is that the cross sections for the nd3/2 states
represent primarilyfinestructure changing collisions, the lone nd
depopulation point (13d) represents depopulation of the mixture of
13dfinestructure states to other levels. Note that this cross section is afactor
of5 smaller than the fine structure changing collision. In a more extensive set
of measurements, Tarn etal.49 measured the ratio of the fine structure
changing to depopulation cross sections for the Cs nd states. Their ratio is
2.4(4) independent of ft, for 6 < n < 10, with the exception of n = 7, which
may be more rapidly depopulated due to an accidental resonance. Precisely
whythe alkali atomsexhibit self quenching behavior so different from the rare
gases is not fully understood, although it is no doubt due to the longer range
interaction of the alkali atom with the Rydberg electron, and perhaps the
Rydberg ion core aswell.

n changing by molecules 221 20 n* Fig. 11.12 Plot of the depopulation cross

sections of the Rb n�states by ground state Rb: rcf(#) (ref. 32);ns (O),
rcd3/2(x), nd (�) (ref. 57);np (D) (ref. 26). The cross sections are plotted vs
effective quantum number n . For reference the geometric cross section
(5/2)jtn4a02 is also shown ( ). n changing by molecules When Rydberg atoms
collide with molecules, the vibrational and rotational degrees of freedom of
the molecule allow the molecule to absorb or release internal energy in the
collision. In addition, the fact that the rotational and vibrational states are
energetically accessible introduces interactions of longer range than the
polarization and short range interaction of the Rydberg electron with an
atomicperturber. Specifically, the electron caninteract withthe multipole
moments of the molecule. For ahomonuclear diatomic molecule, such asN2,
the e~-quadrupole interaction is the longest range interaction, while for a
polar
222 Rydberg atoms molecule such as CO, HF, or NH3 the dominant
interaction is the e"-dipole interaction. Typically it is the rotational dipole
moment which is important. However, it is possible to observe the
interaction between the Rydberg electron and the dipole moment of infrared
active vibrational transitions. The only nonpolar molecule which has been
studied extensively is N2. As we have already noted, i mixing by N2 is not
different from �mixing by Ar or CO. However, n changing is quite different.
The crosssections for depopulation of the Na ns, n<10, states by N2 have
been measured both by time and wavelength resolved laser induced
fluorescence and by sensitized fluorescence in Hg-Na-N2mixtures.24'58'59
The depopulation cross section of the Na ns states section isroughly constant
at ~90 A2 up to the 8sstate at which point it begins to increase slightly.24'58
This cross section is very different from the rare gas depopulation
crosssectionsfor the samelown Nans states. For example the cross sections
for the depopulation of the 6s and 7sstates by Ar are <0.12 A2 and 0.48 A2
respectively.52 This difference is directly attributable to the fact that in
collisions with rare gas atoms energy lost by the Rydberg electron must go
into translation, while in collisions with N2 it can go into vibration and
rotation. The N2 depopulation cross sections of Na ns states of n < 10can be
described with a variant of the molecular curve crossing model used by
Bauer et al.60 to describe the depopulation of the Na 3p state by thermal
collisions with N2. At higher n we would expect that a better description of
Rydberg atom-N2 scattering would be given by picture based on e~-N2
scattering. While such experiments have not been done with Na, they
havewith Rb,61 and these experiments provide an example of a case in
which the free electron picture does seem to work well. Specifically,
collisions of Rbnsand nd stateswithN2have been studied usingtime
resolvedfieldionization for detection. Toobtain the crosssection the decay of
the adiabatic feature corresponding to the initially populated Rb ns or nd
state was observed as a function of time after laser excitation for different
gas pressures. The results obtained for the Rb nsstates are shown inFig.
11.13. Asshownby Fig 11.13 the cross section islargest, 305 A2, at n = 30and
decreases to 170 A2 at n = 46. These measurements agree with cross
sections based on the elastic scattering of the free electron from the N2.
More than 90% of the computed cross section is due to collisional transfers
from the ns states to the nearby (n-3)�,� > 3 states. These collisions occur
because of the short range e~-N2 interaction. Less than 10% is due to n
changing collisions due to the long range interaction of the electron with the
N2 quadrupole moment to produce A/ = �2 rotational transitions. In other
words, N2 behaves almost like a rare gas atom. In collisions with Rb nd
states the fact that N2 is a molecule becomes more apparent. The Rb nd state
cross sections, measured in the sameway, are smaller, with cross sections of
�150 A2, independent of n for 24 < n < 46. These cross sections can also be
compared to those calculated using the free electron model. Consider the 35d
state as a typical example. The computed cross sections for transfer to the
nearby 35� > 3 and 34� > 3 states by the short range e~-N2

n changing by molecules 223 O 1000 I I I I I I 100 46 ns Fig. 11.13

Quenching crosssectionsfor the Rb ns levelsby N2 (ref. 61)and CO (ref. 62)
vs. the principal quantum number n. Experimental values (x). The smooth
curve represents the calculated �mixingcrosssections byN2. The
�mixingcrosssection byCO, whichis not shown in thefigure,is afactor of 2
larger. The triangles (A) represent the total calculated cross sections
including � mixing, n changing and ionization. For CO the line joining the
triangles (A) is drawn for clarity (from ref. 61). interaction are 62 and 18
A2respectively. These crosssections arefar smaller than the measured cross
section of 130(25) A2. The long range e~-N2 interactions leading to the
rotational excitation and deexcitation ofN2 lead to crosssections of 12,5,
and4A2for excitation, deexcitation andionization oftheinitially populated 35d
Rydberg state. The total computed cross section of 99 A2 isstill substantially
belowthe measured value,but itis clear that the short rangeinteractions
arebyno means solely responsible for the observed cross sections. It is also
instructive to compare the n dependences of the depopulation cross sections
of the Rb ns and nd states. The nd cross sections are smaller and n
independent while the ns cross sections are larger and decrease with n. From
the discussion of the theory we expect that crosssectionsduetothe short range
e~-N2

224 Rydberg atoms interactions should decreasewithn

whereasthoseduetolongerrange interactions should not. This expectation also
suggests that the ns, but not the ml, depopulation is due to the short range
interaction. If weconsider apolar molecule, such asCO,weintroduce the
possibility of the longer range electron-dipole coupling, leading to
collisionsinwhich the molecule undergoes A/ = �1 transitions. The
similarity of theNand�mixingcrosssections for N2 and CO shows the
similarity of the short range e~-N2 and e~-CO interactions.45 Unfortunately,
no measurements of n changing by CO have been made at low n. However,
measurements of the Rb ns and nd depopulation cross sections by CO have
been carried out by Petitjean et al., using the same field ionization techniques
described abovefor Rb-N2collisions.62Thetotal depopulation was measured
by observing the decay with time of the adiabatic field ionization feature of
the initially excited state. The crosssections for the depopulation of the Rb ns
states are much larger than the corresponding cross sections for N2. As
shown by Fig. 11.13, the peak cross section is 2000 A2 atn = 25, and the
crosssection decreasesto �1000 A2 atn = 45. An experimental indication of
the reason for the huge difference from N2 is obtained byaqualitative
inspection of thefieldionization signals.Withraregases, and presumably with
N2, there is negligible signal observed at field ionization voltages lower than
the onset of the adiabatic signal from the state populated by the laser. In
contrast, with CO added there is a very obvious, approximately continuous,
signal from zero field up to the threshold ionization field of the initially
populated state. This signal has been interpreted as the diabatic field
ionization signal from higher n states, but whether it is diabatic or adiabaic
ionization, itis clear that a significant amount ofpopulation is
beingtransferred to high n states. Using the free electron model the cross
sections are readily calculated. The short range e~-CO interaction leads to
the ns�> (n�3)�, � > 3transition, which has a calculated cross section a
factor of 2-5 smaller than the observed cross section. The long range e~-CO
interaction, principally the electron-dipole interaction, which leads to/�> /
� 1 transitions in CO,provides most of the cross section, as shown by Fig.
11.13. For substantial cross sections these transitions must be resonant
energy transfers between the CO and the Rydberg electron, i.e. no energy
goes into translation. Rotational deexcitation of the molecule is more
favorable for two reasons. First, the densely packed higher n states make the
resonance more likely. Second, rotational excitation of the CO requires that
energy be removed from the Rydberg electron, and this process is less likely
to occur near its outer turning point than at smaller orbital radii. Although the
intrinsic width of the resonances is computed to be ~1 cm"1, the
manifestations of the resonant character are not particularly striking in the
cross section for two reasons. First, at room temperature there are about 30
rotational states of CO populated, all of which have different resonance
frequencies. Second, there is a substantial ns �> (n�3)�, � > 3
component to the

n changing by molecules 225 cross section, which makes observing small

variations due to resonances more difficult. A most interesting aspect of Fig.
11.13 is that the total cross section declines as n increasesfrom 25
to40,reflecting the decrease incross section for the transitions from the ns to
(n�3)�, � > 3 states. These transitions are due to the short range e~-CO
interaction. The depopulation cross sections of the Rb nd states of 25 < n <
40 are �1000 A2, which isthe same asthe cross section of the Rb ns state if
the ns�> (n � 3)�, � > 3 contribution issubtracted. For the Rb ndstates
the calculated contribution of the scattering of the nd state to nt > 3 and
(n�1)� > 3 stateswithno change in the rotational state of the CO is <100
A2, so 90% of the cross section is due to the inelastic transitions leading to
rotational excitation. Presumably it is because the resonant transfer accounts
for 90% ofthe observed cross section that the structure in the cross section is
more visible in the nd cross sections than in the ns cross sections. For both
the ns and nd states minimal collisional ionization is observed and calculated
inthisnrange,principally because there are too few CO molecules
withenergetic enough A/ = -1 rotational transitions. For example,onlyCO/ >
18 states can ionize an n = 42 Rydberg state by a A/ = -1 transition, and only
3% of the rotational population distribution is composed of / > 18 states.
When molecules such as HBr, or NH3 are used instead of CO there are two
important changes. First, due to the larger dipole moments ofthese molecules
the electron-dipole interaction has a longer range, and the collisions are of
longer duration. Second, due to the larger rotational constant of these H
bearing molecules, the rotational transition frequencies are higher and more
widely spaced. These changes have two effects. First they make it possible
to observe explicitly resonances in the transfer of energy from molecular
rotation to electronic energy of the Rydberg atom. Second, they make
collisional ionization possible. Both these effects havebeen observed
incollisionsbetween Xe Rydberg atoms and NH3.63'64 The experiments
were done using selective fieldionization, and there are typically three
features in thefieldionization spectrum subsequent to collisions; an adiabatic
peak corresponding to the initially populated Xe ni state, a diabatic peak
corresponding to the Xe � > 3 states of the same n, and resolved diabatic
peaks corresponding to specific higher n states. The most striking feature of
these experiments is the clearlyvisible rotational to electronic energy transfer
first observed by Smith et al.63 Their field ionization data showing the final
Rydberg states of Xe 27f-NH3 collisions, shown in Fig. 11.14, exhibit clear
peaks due to diabaticfieldionization of higher n states lying above the initial
n=27 level by the energies of the /�> /-I rotational transitions, for / < 6.
Note that the lowest/ transition, J=2�> 1,leadsto adiabatic ionization
fieldhigher than the adiabatic ionization field for the initially populated Xe
27f state. Kellert et al.64 both measured the total depopulation rates of the
initially populated Xe ni states and analyzed the final bound states resulting
from these collisions. The total depopulation rates of the Xe ni states were
measured by

226 Rydberg atoms CO O * ' � -5 o o ^_ CD CL CO Hd O o (a) (b) No

Xe(27f) -10 Torr NH3 , 7/xsec delay co ^> ^" ^o C\J A A A A A A 1^- CO
IO ^ rO OJ I I I 1 1 I I 1 I I 0 500 1000 Field strength,V/cm Fig. 11.14
Fieldionization signalwhentheXe 27f state is populated as afunction
ofionizing field strength (a) without NH3, (b) with 10"5 Torr NH3. The
resonant rotational NH3 transitions are indicated beneath eachfieldionization
feature (from ref. 63). monitoring the dependence of the adiabatic field
ionization signal from the nf states as a function of time after the laser
excitation for different NH3 pressures. For n = 25-40 the rate constants are
~2 x 10~6 cm3/s. Analysis of thefinalstates under single collision conditions,
shortly after the laser excitation showsthat, not surprisingly, the depopulation
is dominated by � mixing collisions, but that the resonant n changing
collisions to higher n states comprise �1/3 of the total depopulation rate. At
long times after the laser excitation itis clear that the atoms transferred to the
high � states will contribute to the resonant transfer signals shown in Fig.
11.14. Since the high � states are, for all practical purposes,

n changing by molecules 227 degenerate withtheni statestheir participation

doesnot affect thelocations of the resonances of Fig. 11.14. The Xe ni states
are nearly degenerate with the nt > 3 states. The Rb ns states, however, are
more isolated, having aquantum defect of 3.13. Kalamarides etal.65 studied
collisionsof Rbns atoms,40 < n < 48, withHFandobserved that thens�� (n-
3)�, � > 3 rate constants in this case only comprise 30% of the total
depopulation rate constants, which are 10(2) x 10~7 cnrVs for 40 < n < 48.
Roughly 60% is dueto the n changing (An >0)transitions resonant with/ =
1�> 0 rotational transition in HF, and 10% is due to n changing, An < 0,
transitions resonant with the HF / = 0 �� 1 transition. Although we have
discussed n changing collisions first, ionization of He, Ne, and Ar Rydberg
atoms bymoleculeswas observed before nchanging collisions by Hotop and
Niehaus,66 who, not having selective methods of preparing or detecting the
Rydberg states, could not detect changes in n, but only ionization. They
observed ionization cross sections of ~104 A2 for ionization by H2O, NH3,
SO2, C2H5OH, and SF6, noting that in the caseof SF6the production of a
positive ion led also to the production of SF6~. They were not able to detect
ionization in collisions with the molecules H2, O2, N2, NO, or CH4. Of
these only NO is polar and has dipole allowed rotational transitions, but the
dipole moment is small. More important, the NO rotational constant is small,
and only the highest rotational states can ionize a Rydberg atom by a AJ = -1
rotational transition. The situation is the same as for the collisions with CO
described earlier. The fact that only rapidly rotating polar molecules caused
collisional ionization was also demonstrated by Kocher and Shepard,67'68
who showed that this same class of molecules led to collisions with large
changes in n aswell. A good example of a rapidly rotating molecule is NH3,
and, as shown in Fig. 11.14, NH3 molecules in states of / > 7 can ionize Xe
27f atoms by a AJ = -1 transition. If we goto higher n states lower rotational
states of NH3can ionize the initially populated state, and the ionization rate
should increase accordingly. Kellert etal.64measured theionization ratesin
twoways. Oneof thesewas touse a lowfieldpulsetocollecttheXe+
ionswhichhadbeen formed inshortenoughtimes after laser excitation, �5/^s,
that singlecollision conditions prevailed. Comparing the number of ions
produced to the number of atoms remaining in the initially populated ni state
gives the cross section, or rate constant. This method is, in essence,
thesameastheoneusedtodeterminehowmuchofthetotal depopulation
crosssection is due to � mixing asopposed to n changing collisions.The
measured rate constants of Kellert etal.64for ionization and total
depopulation are shown in Fig. 11.15 along with the calculated ionization
rate constants11'69'70 of the ni state. While measured and calculated
ionization rate constants have qualitatively similar behavior, the measured
rate constants are roughly a factor of 4 larger than the calculated ones. An
interesting aspect of Fig. 11.15 is the step structure in the calculated
ionization rate constant, reflecting the fact that as n increases lower rotational
states are added to the number capable of ionization.

228 Rydberg atoms io5 ro o < 7) O O -7 io7 IO8 /GEOMETRIC � |0

CROSS SECTION " nil 10" io" � O LJ CO CO GO O a: o io-12 20 25 30
35 40 PRINCIPAL QUANTUM NUMBER Fig. 11.15 Principal quantum
number dependence of the experimental rate constants for the total
depopulation of the Xe ni states by NH3 (�) and for collisional ionization
(O). Also shown isthe calculated ionization rate constants of Rundel (R) (ref.
69),Latimer (L) (ref. 10), and Matsuzawa (M) (ref. 70) (from ref. 64). The
data of Fig. 11.15 are too sparse to reveal such structure, but it has been
observed in collisions between Kr Rydberg atoms and HF, which has a
larger rotational constant than NH3.71 As we shall see shortly SF6 readily
removes the electron from a Rydberg atom and is hence an excellent detector
of Rydberg atoms. Using first a mixture of SF6 and Kr, Matsuzawa and
Chupka scanned the wavelength of a vacuum ultraviolet source across the Kr
Rydberg series and observed the SF6~ ions produced from the Rydberg
atoms, which remain in the observation volume for microseconds.71 Below
the ionization limit they observed the Kr Rydberg series, and above the limit
they observed no signal since the

n changing by molecules 229 400 10 11 Fig. 11.16Total cross sections ans

for depopulation of the Na ns states by CH4 (�, ) and CD4 (A, ). The
smooth dotted curve (....) shows the expected depopulation cross section in
the absence of resonant e~-v transfer (from ref. 72). photoelectrons rapidly
leave the observation region. When HF was added the SF6" signalfrom the
high nstateswas substantially depressed. Asthe wavelength is increased from
the ionization limit the SF6" signal can be seen to increase in steps
coincident with the energies of the HFJ-* J - 1 transitions. The clear
evidence for resonant transfer of energy from molecular rotation to Rydberg
atoms prompts us to ask if an analogous process might occur between
molecular vibration and Rydberg atoms. In general, vibrational frequencies
are higher than rotational frequencies, so we would expect to see evidence
for resonant transfer at relatively low n, <10, and, in the vicinity of room
temperature, we would expect to seeenergy transferred from the atom to the
molecule. In the depopulation of theNans statesbyN2there is nosignof any
resonant transfer, the depopulation cross section increases monotonically
with n, consistent with a curve crossing model.24 On the other hand, laser
induced fluorescence measurementsof the depopulation crosssections of
theNans statesbymethane, CH4, and deuterated methane, CD4, show
evidence of resonant behavior similar to that seen for Xe nf states with
NH3.72 In Fig. 11.16 we show the depopulation cross sections of the Na ns
states by CH4 and CD4. In Fig. 11.16 the dotted line shows

230 Rydberg atoms Table 11.1. Observed resonant transfersfrom Na

tomethane and deuterated methane.a Na Na frequency transition (cm"1)
Molecule Vibrational mode, rotational branch Branch center frequency
(cm"1) Branch width (cm"1) 5s-> 6s -> 7s -> 4p 5p 5d 2930 1331 975 CH4
CD4 CD4 v3,P v4,R v4,R 2940 1340 965 60 60 50 "(fromref. 71) the
"resonance free" behavior, asseenwithN2. It is evident that the 5sand 6s CH4
cross sections and the 7s CD4 cross section lie above the smooth dotted
curve. The increased cross sections for these three states are attributed to
resonant electronic to vibrational energy transfer. Table 11.1 identifies the
three atomic transitions and the resonant molecular transitions in CH4 and
CD4. For example the rapid depopulation of the Na 7s state by CD4 is
attributed to the Na 7s�> 5d transition. To verify this assignment the cross
section for the 7s�� 5dtransfer was measured for both CH4and CD4 by
observing the 5d-3p fluorescence as well as the 7s-3pfluorescence.The
7s�> 5d cross sections are 215 A2 for CD4 and 15A2 for CH4. As shown
by Fig. 11.16, the 7s CD4 cross sections is �240 A2 above the smooth
dotted curve in good agreement with the 7s�> 5d cross section. Similar
confirmations were carried out for the other two resonant collisional
transfers. The observed resonant energy transfer cross sections can be
described in terms of afree electron scattering from the CH4or CD4. First,
both the v3 and vA modes are infrared active, and there is a long range
electron-dipole interaction.73"75 Second, electron scattering measurements
have shown that the cross section for electron impact excitation of the v3 and
v4modesis high atthreshold.76 As aresult of the second point the Rydberg
electron is able to excite the methane at the largest orbital radius at which it
isenergetically possible, i.e. just within the outer turning point of thefinalNa
states. Electron attachment Collisions of some halogen bearing molecules,
such as SF6, with Rydberg atoms result in attachment of the Rydberg electron
to the molecule to form a negative molecular ion. Simple attachment,
dissociative attachment, and attachment followed by autodetachment have all
been observed. Collisions with attaching molecules are an excellent example
of a process dominated by the electron

Electron attachment 231 molecule interaction, for at high n the rate constants
for attachment of the electrons in a Rydberg atoms are identical to those for
attachment of free electrons, as predicted by Eq. (11.32).77'78However, at
low n the rate constants fall below the free electron value. Measurements of
the cross sections for attachment are most often done using time resolved
selective field ionization. In collisions of Rydberg states of n > 20 with
attaching targets no n changing collisions have been observed, but � mixing
collisions have.79Often they are negligible,but even the largest reported
�mixing rates, for the Xe ni states with molecules, are �30% of the
collisional attachment rates. The most straightforward way of measuring the
attachment rate is to observe the decay of the sumof thepopulations
intheinitiallypopulated state and those populated by �mixing for short times
after laser excitation. By short times we mean those during which there is not
much depopulation of the initially populated state. At short times this decay
rate, y, is approximately79 y = yo; + 7att, (11.42) but at long times it
becomes y = yo� + 7att, (n.43) where yOi and y0� are the radiative decay
rates of the initially populated state and the mixture of �states respectively.
Measuring the decay rate as a function of the molecular gas pressure gives
the attachment rate constant since � mixing and ionization are the only two
significant processes. In Fig. 11.17 are shown the free electron attachment
cross sections for a series of halogen bearing molecules derived from
Rydberg atom measurements78 and swarm measurements.80 The
crosssections ofFig.11.17 areobtained from theXeniRydberg statesof25< n <
40using a variant of Eq (11.32), aatt = � (11.44) Vrms where k isthe
measured attachment rate constant and vrms isthe rms value of the electron's
velocity, which scales as \ln. Weuse the fact that the rate constants for
attachment of a Rydberg electron and a free electron of the same average rms
velocity are the same, as shown in Eq. (11.32). In Fig. 11.17 the most striking
feature is the continuity of the cross sections from the swarm measurements at
higher energies80 to the Rydberg atom measurements at low energies. This
continuity, in essence, verifies the applicability of Eq. (11.44). As noted
above, attachment can occur in several ways, and analyzing the negatively
charged products allowsusto differentiate between them. Specifically,
negatively charged products can be identified by theirflighttimes to a
detector.79 Representative examples of attachment processes are shown in
Fig. 11.18, the time resolved ion signals subsequent to the collisions of Xe
26f atoms with attaching targets in the presence of a 2 V/cmfieldinthe
interaction region. For all three molecules shown in Fig. 11.18 the Xe+
signal exhibits the expected

232 Rydberg atoms io-13 io 6 8 10 20 40 60 80 ELECTRON ENERGY

(meV) Fig. 11.17 Cross sections for Rydberg electron (�) (ref. 78) and free
electron (O) (ref. 80) attachment to C5F8, 1,1,1-C2C13F3, CC13F,SF6, and
CC14. The Rydberg electron attachment cross sections are calculated using
Eq. (11.42) (from ref. 78). exponential decay. The temporal displacement of
the beginning of the ion signal from the laser pulse is due to the Xe+ flight
time to the detector. When Xe collides with CH3I, to produce I~, the flight
time is consistent with that expected for I", shown by the arrow. It isnot
inconsistent with the flight time of CH3I", so the most stringent statement that
can be made on the basis of this measurement alone is that the negatively
charged particle contains I". The assignment as I"is due to the

Electron attachment 233 arbitrar y uni i JNE L ( ER CHAl v COUNT S P

�� i ' � � 4 ^ Xe|25f>*CjF|4 / . i . � t-.c7F,; * � � * *^ Xe|25f>
Xe|25f> � / 1 � CTFI4 CHjl V � Xe4 � r 10 15 20 25 30 35 40 T !ME
AFTER LASER PULSE (/xsec) Fig. 11.18 Arrival time spectra of the
products of the collisional ionization of Xe 26f high Rydberg atoms by CjFu,
CH3I, and C6F6. As shown, collisions with CH3 lead only to r .
Collisionswith C7F14lead toboth C7F14~ and e", asshownbythe large signal
atearlytimes due to electrons. C6F6leads to the production of alonglived
autodetaching state of C6F6" which produces a nearly continuous electron
signal at early times (from ref. 79). higherresolution mass spectrometricwork
of Stockdale etal.81 who only observed I~ when free electrons were
attached to CH3I. In contrast to CH3I, CyF^ produces both electrons and
C7F14~. The electrons have essentially zero transit time on the scale of Fig.
11.18, and the QF^" has a transit time of 25 jus. Experimentally itis not clear
ifthe electron signal comes from simple collisional ionization or collisional
attachment followed by very rapid autodetachment. In Fig. 11.18 the electron
and C7F14" signals are comparable in size. Although the relative sensitivity
of the detector to electrons and C7F14" isnotknown, if we assume the
sensitivities are equal, implying that the two processes have comparable
rates, we can reconcile the fact that the rate constant for Xe+ production from
Rydberg states exceeds the attachment rate constant obtained from the swarm
experiments of Davis etal.82 by afactor of2-3. Finally, inFig. 11.18 Xe 26f
atoms colliding with C^6 leads to a continuous signal followed by a more or
less

234 Rydberg atoms g iol3 O jj IT) r CROS S 5 T- -14 AT TA D O _ - $

Rb(nd) { Rb(ns) 5 Xe(nf) 2 A SWARM DATA i i i i i i i i 1 i TPSA/ 1111 I
� S ^ KLOTS i i i 1111 10 100 ELECTRON ENERGY (meV) Fig. 11.19
Cross sections forRydberg electron, Rb nd (�),Rb m (x) (ref.84), Xe ni(D)
(ref. 85),and free-electron (A) (ref. 88) attachment to SF6. The solid lines
show the theoretical result of Klots (ref. 86) and the results of threshold
photoelectron spectroscopy by electron attachment(TPSA) studies (ref.87)
(from ref. 84). normal exponential decay starting 18 jus after the laser pulse.
The normal decay starting at 18JUS isdue to the detection of C6F6~. The
earlier signalisattributed to autodetachment electrons from long lived C6F6~
states. This assignment is consistent with previous observations of Naff
eta/.83 that C6F6~~ produced by attachment oflow energy electrons has a
12JUS autodetachment lifetime. One ofthe reasons for using Rydberg atoms is
that itis in principle possibleto observe electron scattering processes at very
low electron energies. An elegant example of thisnotion isthe extension of the
measurements of attachment by SF6 to very high n, byZollars et al.S4 They
used the inherently high resolution of a single mode cwdye laser to resolve
the high lying, n > 100,Rydberg states of Rb by two photon excitation. Their
measured rate constants are approximately the same, 4 x 10"7 cm3/s, as the
rate constants forlower n states.85 These measurements agree with
calculated86 andmeasured87'88 free electron attachment rate constants. The
nindependence of the rate constant indicatesthat the free electron attachment
crosssection scalesas theinverse of theelectron velocity, asexpected.
Thispoint is shown explicitly byFig. 11.19,a plot of the crosssection asa
function of the electron energy. Asshown byFig. 11.19, themeasurements
godown to electron energies of 1 meV, an energy unattainable by
conventional means. At lown the rate constants for free electron attachment
and attachment of Rydberg electrons areno longer the same. In some cases a
dependence of the attachment rate constant onthe �ofthe Rydberg
statehasbeen observed,89'90 and at low n the rate constant falls below the
free electron value.90"92 For

Electron attachment 235 example, the rate constant for attachment of the
electron in Rydberg Knd states by SF691 falls precipitously from its
asymptotic value of 4 x 10~7 cm3/s below n = 20. There are two reasons for
the discrepancy between the free electron attachment rate constant and the
Rydberg atom rate constant at lown. Thefirstis that the attachment rate
constant is n independent only in the limit that the attachment cross section is
small compared to the geometric cross section of the Rydberg atom. In other
words on any given passage of the attaching molecule through the Rydberg
atom the probability of attachment is assumed to be much less than 1. When n
decreases to the point that the geometric cross section approaches the high n
attachment cross section, the attachment cross section and rate constant fall.
A second effect which should further suppress attachment at low n is that the
negative molecular ion may not have enough energy to escape from the
positive ion of the Rydberg atom.91 If the molecule captures the electron at a
distance R from the Rydberg ion core, it must then overcome the attractive
coulomb potential �1/R to escape. As a result, free positive and negative
ions are only possible if the relative velocity of the Rydberg ion and the
molecule isenough that the relative translational energy exceeds 1/R. If this
condition is not met, attachment can occur, but the result is an orbiting pair of
ions which most likely recombine to produce deactivation of the initially
excited Rydberg state or a chemical reaction with neutral products. This
point is made by Fig. 11.20.93 For high n the K nd states have total
depopulation and attachment rate constants which are very nearly the same,
but low n K nd states have depopulation rate constants substantially larger
than the negative ion formation rate constants,91'93 suggesting that temporary
attachment mayoccur, but coulomb trapping prohibits the formation of afree
negative ion.93The calculated rate constant ofFig. 11.20 is obtained
byrequiring that the relative velocity of the colliding atom and molecule
exceed 1/R. For a thermal velocity distribution this requirement leads to a
rate constant for negative ion formation scaling as e~w/kT where W is the
binding energy of the Rydberg state.91'93 Beterov etal.92 have measured the
electron attachment rate constants for Na Aip-SF6collisions, extending the
measurements to slightly lower values of n than those of Fig. 11.20.By
comparing SF6~ production to black body photoionization Harth et al.90
have measured the rate constants for Ne ns, nd-SF6 electron attachment
collisions down to effective quantum number n* = 5. The measured Rydberg
atom-SF6rate constants are shown in Fig. 11.21.90At high n* the results
areinsubstantial agreement witheach other and the free electron attachment
rate constant. At low n* the rate constant does not exhibit an ew/kT scaling.
Such a scaling would lead to unobservably small rate constants for n < 10.
Two suggestions have been advanced as to why the rate constant does not
vanish at low n. First, Beterov etal92 have suggested that before the electron
is captured by the SF6 n increasing collisions with SF6 lower the binding
energy to the point that coulomb trapping does not eliminate SF6~
production. In their

236 Rydberg atoms htife* 10 15 20 25 PRINCIPAL QUANTUM NUMBER n

Fig. 11.20 The n dependence of the rate constants for collisional
depopulation of K nd atoms by SF6(�) (ref. 93)and for formation offree
SF6" ionsincollisionsof SF6withKnd (A) (ref. 93), K nd (A) (ref. 91), Na
np(#) (ref. 92), Xe ni (�) (ref. 85). Also shown are rate constants calculated
with (- - -) and without ( ) taking into account postattachment electrostatic
interactions (from ref. 93). picture the SF6 is treated as being structureless.
Using this model Harth et al90 have been able to reproduce the rate constants
of Fig. 11.21. Second, some of the internal energy of the SF6" can be
converted into translational energy after the electron is captured. If a constant
fraction of the available internal energy is assumed to be converted to
translational energy, it isnot possible to reproduce the n* dependence of the
rate constant shown in Fig. 11.21. However, assuming a lM*4 scaling of the
fraction of internal SF6" energy transferred Harth etal.90 have been able to
match their experimental rate constants over the entire range of Fig. 11.21.
The 1/n*4 scaling of the energy transfer fraction implies that the transfer
occurs more readily for low than high Rydberg states. This notion is
supported by the work of Harth et al. who observed CS2~ resulting from Ne
fts, ftd collisions with CS2.90 At high ft the rate constant for the production
of Ne+ is a constant, indicating that the Rydberg electron is easily removed
from the Ne. In contrast, the rate constant for CS2~ production peaks at ft* =
18, not at high ft*, asshown by Fig. 11.22. Thisunusual n dependence of the
rate constant suggests that only when the CS2~ is formed near enough to the
residual Ne+ can it dispose of enough internal energy to live long enough,
35jus, to be detected.90 A graphic illustration of the effects of the interaction
of the two ions is obtained by observing the spatial location of the parent
Xe+ ion after attachment by SF6.84

Electron attachment 237 IF 10 6 i \J 10-7 10"8 10-9 1o-io 10"11 : 1 ' ' ; % - ?
A(nf) + SF61 1 | 1 1 T 1 | 1 1 1 o � � i oO��# A %J<* f O Jr , �A+ +
SF6" i I i i i i I i i i i 1 * * * * 1 ; � = - z - z 10 15 20 25 30 Fig. 11.21 Rate
constants for SF6 formation in collisions of Ne, K, and Na Rydberg atoms
with SF6 vs effective quantum number �*, Ne ns (O), Ne nd (�) (ref.90), K
nd (A) (ref.91), and Na np (�) (ref. 92) (from ref. 90). A beam ofmetastable
Xe atomsis excited to aRydberg state or photoionized and the product Xe+
ions are expelled by afieldpulse from 2-5JUS later andimpinge upon a
position sensitive detector. When the metastable Xeisphotoionized the pattern
reflects the geometry of the metastable Xe beam, and the same is true when
Xe+ isproduced by attachment of the electron from aXe 60f state toSF6.
Apparently, there is no deflection of the Xe+ by the SF6~. However, atlow
n,n< 40, the pattern of Xe+ detected is twice as broad as the metastable
beam, indicating deflection of the Xe+ion bythe SF6~. Asexpected,
thedeflections of the Xe+ become more pronounced at lower n. If we
consider dissociative attachment, the dissociation of the molecular ion
provides translational energy to the fragments, and thisenergy mayovercome
the coulomb trapping. Walter etal.94 studied the attachment ofthe electron of
ahigh n, n ~ 55, Rydberg atom to CH3I to form CH3I~, which dissociates
rapidly, yielding I~. They found that most ofthe available energy, �1/2 eV,
from forming the negative ion went into translational energy of the molecular
fragments CH3 and I". Byobserving the spatial distribution ofI~ from
dissociative attachmentof electrons in K nd Rydberg states by CH3I atlown
Kalamarides etal.95 observed a veryclear illustration of
coulombtrapping.The K atomsinabeam areexcited bya laser asshown byFig.
11.23, andthe I~ ions aredetected byaposition sensitive detector. When the K
26d state is excited the I~ distribution is apparently isotropic, asshown
byFig. 11.23. Asn islowered to 15and9the I~ distribution

238 Rydberg atoms |ur710'12 e 10"15 l 0 -16 10-17 b 7 10 1 1 10 I 15 1 ' 1

15 20 i i I i 20 . | .. . 25 1111 25 i 111 30 d 30 s 10 15 20 25 30 n* Fig. 11.22
Rate constants k and cross sections o for CS2 formation in collisions of Nens
(O) andnd (�) Rydberg atomswithCS2 moleculesvs n*.Thedecreaseof
therate constant and cross section for n* > 20 is due to the reduced
probability of stabilizing Ne+ encounters (from ref. 90). becomes more
sharply peaked in the direction opposite the K beam's. This angular
distribution is a reflection of the fact that atn = 9onlyifthedissociation of
CH3~ produces I~ moving in the direction opposite to the K+ motion is there
enough relative energy for the K+ and I" to overcome the coulomb barrier
and separate. The use of theweakly bound electron in a Rydberg atom to
measure lowenergy electron attachment rate constants has proven to be one
of the more useful applications of Rydberg atoms. Measurements have been
refined to the point of measuring the lifetimes of negative ions formed by
attachment,96 and it is likely that further developments will follow.

Associative ionization 239 K I 67-100% BEAM PSD 34-67% 1-34% 4 V

./LAS ^ ~ r^T^* - - ^ D IT A LASER BEAM n= 15 Fig. 11.23 Spatial
distribution of I ions detected following K nd-CF3I collisions at intermediate
n. The inset indicates the directions of the laser and potassium atom beams
(fromref. 95). Associative ionization Associative ionization is
theprocessinwhich anexcited atomcollideswitheithera ground state atom or
molecule to form a molecular ion and a free electron. For example, the
associative ionization of aKRydberg atom bycollision partner M is
represented by Kn( + M ^ KM+ + e~. (11.45) Ignoring the translational
kinetic energy of the collision partners, associative ionization isin fact the
only way in which a collision with aground state atom can ionize a Rydberg
atom. If weconsider collisionsbetween alkali Rydberg atomsand ground state
atoms, associative ionization is possible as long as the energy of the Rydberg
state exceedsthe minimum energy of the dimer ion, as measured from the
energy of the ground states of two separated atoms. This notion is shown
graphically in Fig. 11.24, and Lee and Mahan97 used this requirement to
measure the well depth of the alkali dimer ions. Using a lamp and a
monochromator they excited the np levels of alkali atoms in a cell and
measured the monomer and dimer ions produced. Theywere ableto
discriminate between, for example, Cs+ and Cs2+ by their mobilities in Cs
vapor (Cs+ has a lower mobility due to resonant charge transfer). They
observed that when nplevelsof n < 12were excited, Cs2+ was the dominant
ion, whereas above n = 12,Cs+ was dominant, indicating that below n = 12
associative ionization is by far the dominant ionization mechanism.
240 Rydberg atoms <r LLJ LU Rb 5s + 5s Fig. 11.24 Adiabatic
potentialcurves for the Rb-Rb system. The initially populated Rb nt
Rydbergstatecanleadto associativeionizationifitsenergyatR = ��
exceedstheminimum of theRb+ + Rb5s potential well,as shown. Associative
ionization was also observed between atoms in Sr 5sn( states and the ground
Sr 5s2 state in a beam of Sr by Worden etal.98They used multi step resonant
laser excitation of the Sr 5sn( states, and at Sr number densities of 1013
cm"3, theyobserved Sr2+. Theywere ableto determine thewelldepth of Sr2+
tobe 0.77 eV. Systematic measurements of associative ionization have been
done in Na,99'100 Rb,101102 and Ne,103byseveral different methods.
Measurements of the associative ionization rate constants of high lyingNa nt
states were made using abeamof Na, which was excited by two pulsed dye
lasers from the ground state to the 3p state and then to an n� state.99 Stark
switching allowed the population of the np states and amixture of �>
2states. Theexcited atomswereallowed to collidefor5 jus after which the Na+
and Na2+ products were accelerated out of the interaction region toward the
detector by a pulsed extraction field. Na+ and Na2+ were distinguished by
their flight times to the detector. To determine the associative ionization rate
constant, the densities of the ground and n� Rydberg states must be known,
as well as the Na2+ production. The number of Rydberg atoms was
determined by field ionizing the Rydberg atoms on alternate shots of the
laser. Thenumber of Rydberg atoms,together withthelaserbeamgeometry,
gives their density. The density of the 3s atomswas determined bymeasuring
the lengthening of the 3p lifetime due to radiation trapping. Specifically, the
variation in the population of the Rydberg state was measured as a function
of the time delay between the two lasers, a method which works well for Na
densities in excessof 1010cm"3.The results of the experimental
measurements of associative ionization of Na Rydberg state atomswithground
state Na atoms areshowninFig. 11.25.As

Associative ionization 241 iL 5 10 15 20 25 n* Fig. 11.25Experimental and

theoretical rate constant for associative ionization of Nani + Na 3satoms at T
= 1000 K plotted against effective principal quantum number. Results for 17<
n* < 27(ref. 99) are scaled to absolute results obtained byBoulmer etal. for
5< n < 15 (ref. 100). Experimental data;/is levels(�);/iplevels (�);��
>2levels (O). Theoretical curves: associative ionization of nslevels ( ), np
levels ( ), ni > 2levels (� o �), associative plus Penning ionization ( )
(from ref.99). shown, the rate constant, or the cross section, peaks at n = 11
and decreases at higher n. Extensive experiments based on rather different
methods were used to study the Rb n( + Rb 5s associative ionization.
Klucharev et al.101 used a lamp and a monochromator, an approach not
unlike that of Lee and Mahan,97 and Cheret et al.102 used cw dye lasers.
The results are very similar to those shown in Fig. 11.25, and are in good
agreement with the rate constants calculated by Mihajlov and Janev.104 By
comparing black body photoionization and associative ionization signals
Harth et al.103 measured the associative ionization of Ne ns and nd Rydberg
atoms with both Ne and He ground state atoms. Using Ne they found results
almost identical to those shown in Fig. 11.25, but using He the cross sections
were observed to be almost two orders of magnitude smaller. The alkali
associative ionization cross sections can be understood using a model
proposed by Janev and Mihajlov.105 In Fig. 11.26,we show the potential
curvesof the Rb-Rb system. R is the internuclear separation. At small R the
potentials connecting to the same R = <*> states are split by the exchange
interaction. The initial state of associative ionization is Rb ni + Rb 5s,which
isshown by a broken line. If the two Rb atoms approach to within small
enough R that the exchange splitting is significant, the molecular Rb n�-Rb
5s state converging to the nearby

242 Rydberg atoms Rb++Rb5p Rb ni + Rb 5p Rb+ + Rb 5s "" Rb ni + Rb ns

Fig. 11.26 Rb-Rb potential curvesshowingthe origin of the differing rates for
Penningand associative ionization. In associative ionization theinitial state
Rbn�+ Rb5s onlyis above the lower Sg ionic state at small R where the 2g
� 2U exchange splitting is large. Only at small R does autoionization to the
ionic molecular state occur. In contrast, in Penning ionization the initial state
Rb ni + Rb 5p always lies above the ionic final state, and autoionization can
occur at any R. 2U molecular ion state can autoionize into the degenerate
continuum of the lower 2g molecular ion state. This picture suggests that the
peak cross section should come from the value of R at which the
autoionization rate becomes significant, �10 A2. In other wordsthepeak
crosssection should be �100 A2, whichfor Rb at 450 Kcorresponds to arate
constant of 0.5x10~9 cm3/s, approximately the highest rate constant of Fig.
11.25. The model also implies that the cross section should decrease roughly
as n~3, afactor reflecting the scaling of autoionization rates due to the
normalization of Rydberg electron's wave function at the ion core. As shown
by Fig. 11.25, the rate constants fall sharply at high n. In Fig. 11.26 the
associative ionization ofNa/i�-Na collisionswasexplained intermsof
autoionization from the repulsiveupper 2Umolecular statetothe attractive
lowerSgstate. In the rare gas systems the lower state is a 2 state but the upper
state is a II state, precluding autoionization bythe ejection of an s wave
electron.103 For this reason the Ne rc�-He cross section is quite small, and
one might expect the Ne n�-Ne crosssection to be also. However, for the
Nen�-Ne casethe presence of resonant charge transfer channels enables the
autoionization.103 Associative ionization measurements have been done with
molecules as well. Specifically, associative ionization of K nd states with a
variety of molecules has been studied.106'107 K atoms in a thermal beam
were excited with pulses formed

Penning ionization 243 from a cw laser beam at a high repetition rate, 10

kHz. The K nd atoms were allowed to collidewith a mixture of SF6and, for
example,H2Sfor aperiod of 1-10 jus, after which the positive and negative
ions formed were collected. Using the known rate constants for electron
attachment by SF6, the observed KH2S+ and SF6~ signals immediately
imply the ratio of the H2S associative ionization cross section tothe
SF6electron attachment cross section. Tooffset the much largersize of the
SF6attachment cross section ahigher densityofH2Swas usedthan SF6, with
typical densities 10n-1012 cm"3 and 3 x 1010 cm"3, respectively. The
results of these measurements, carried out for H2S, H2O, CH3OCH3, and
other similar polyatomic molecules, are uniformly small rate constants,
~10~n cm3/s, for the n � 9to n = 15 Knd states. These rate constants are two
ordersof magnitude smaller than the rate constants for associative ionization
of Rydberg atoms with alkali ground state atoms. Why are the cross sections
so small? A possibility suggested by Kalamarides et al.107 is that the dipole
moments of these polar molecules may produce more �mixing of the states
of the Rydberg electron than do alkali atoms. On average, � mixing puts the
Rydberg electron into a higher �state inwhich it is further from the core and
thus less likelyto autoionize, in the terms of the model of Janev and
Mihajlov.105 Penning ionization Penning ionization occurs when the
colliding pair of atoms has enough electronic energy to create a free atomic
ion, i.e. the sum of the electronic energies must exceed the ionization
potential of one of the atoms. The one case which has been studied is Rb n�
+ Rb 5p -^ Rb + + Rb + e~. This system has been studied by Barbier and
Cheret using two step cw laser excitation of Rb in acell coupled with mass
spectroscopic detection of the product ions.108 They obtained the Penning
ionization rate constants shown inFig. 11.27. Comparing Figs. 11.27and
11.25,it isapparent that the Penning ionization rate constants exceed those for
associative ionization bytwo orders ofmagnitude. Thereason for this,pointed
out byBarbier etal.,109 is shown schematically in Fig. 11.26. The entrance
channel for Penning ionization isRb n� + Rb 5p,which is shown asadotted
line just belowthe excited state molecular ion potential curve. At virtually
any value of R, not only R small enough for a significant exchange splitting, it
can undergo the resonant dipoledipole energy exchange Rb n� + Rb 5p-^Rb
s��l+Rb 5s. Ion-perturber collisions While most collisions involving
Rydberg atoms are those in which the perturber interacts primarily withthe
outer electron, insomecasesthe dominant interaction

Rydberg atoms 10" n# Fig. 11.27 Individual Penning rate coefficients for Rb
ni + Rb 5p, kpi(n�), against the effective quantum number n* (from ref.
108). is with the Rydberg ion core. Clear evidence for Rydberg ion-perturber
interaction is the deflection of a beam of Li Rydberg atoms by collisions with
gases observed byKocher and Smith,110 sinceonlyinteraction of theperturber
with the Rydberg ion core can lead to observable deflection of the Li atomic
beam. They excited a thermal beam of Li atoms to Rydberg states using a
pulsed electron beam which delivered 10jus bursts at an 850Hzrepetition
rate. The beam of Rydberg atoms was collimated and passed through 35 cm
of rare gas at a variable pressure. They verified that the rare gasdeflected the
Rydberg atoms out ofthe collimated beam, andbymeasuringthe attenuation
oftheonaxisbeamthey determined the rate constants given in Table 11.2. Since
the Rydberg atom beam waspulsed they could measure the rate constant as
afunction of the atomicbeam velocity and found it to be independent of it,
implying a cross section with a 1/v dependence. By changing the
ionizingfieldin the detector they could change the lower bound of n detected
from 35to 65, and no change in the rate constant was detected. Both the n
independence andthe 1/v dependence ofthe crosssection agreewith a
description based on the perturber's scattering from the Li+ core. Since the
electron is not involved, the cross section is n independent, and the
chargeinduced dipole interaction leads to a 1/v dependence of the cross
section. Furthermore, the measured rate constants are in good agreement with
those computed for the scattering of Li+ from the perturbers, as shown in
Table 11.2.

Fast collisions 245 Table 11.2. Experimental and calculated rate constants
for deflection of Li Rydberg atoms by five target gases.a Rate constant
experimental theoretical Gas (1(T9 cm3/s) (1(T9 cm3/s) He 1.97(40) 1.75
Ne 1.88 (45) 2.42 Ar 2.37 (65) 4.94 H2 4.00 (65) 3.47 N2 2.12(60) 5.12 a
(from ref. 110) For typical thermal velocities, the rate constants of Table
11.2 correspond to cross sections of -100 A2. Another clear illustration of
scattering from the Rydberg ionic core is the isotopic exchange observed by
Boulmer et al.111 In a He afterglow composed of 50% 3 He and 50% 4 He
they populated, using laser excitation from the 2s state, either 3 He or 4 He
np states and observed the fluorescence from both 3 He and 4 He Rydberg
states. They observed that the 300 K rate constant for the process. 3 He n = 9
+ 4 He Is -> 3 He Is + 4 He n = 9 (11.46) is 5.7(10) x 10~10cm3/s. The
results are unchanged if the roles of 3 He and4 He are interchanged. This
process is interpreted as charge exchange between the Rydberg ion core and
the perturbing ground state He atom with the Rydberg electron remaining a
spectator. This interpretation is supported by several observations. First, the
total exchange rate constant, including processes in which there is a change
in n of the Rydberg electron, is only slightly larger, 6.8(20) x 10"1 cm3/s.
Second, the theoretical 300 K ion-atom charge exchange rate constant is 5.3 x
10~10 cm3/s.112'113 Finally, the extrapolation of the experimental114 ion-
atom rate constants leads to good agreement with these observations. Fast
collisions All the collision processes we have discussed in any detail are
thermal collisions. We would now like to return to apoint made early inthe
discussion of the theory. If the collision velocity is high compared to the
Rydberg electron's velocity, the Rydberg atom-perturber cross section should
be equal to the sum of electronperturber and the Rydberg ion-perturber cross
section at the same velocity. A

246 Rydberg atoms -(eV) 2.5 3.0 3.5 8 9 10 II 12 13 WD(keV) Fig. 11.28
Measured cross section for ionization of �* = 46 (35 < n < 50) D atoms in
collisions with N2 vs the kinetic energy WD of the deuterium atom (O),
measured cross section for destruction of n* = 46 D atoms (�) and n* = 71
D atoms (A) in collisionswith N2 (ref. 116).Forcomparison,
Kennerly'smeasured totalcrosssectionfor free electron-N2 scattering vsWe( )
(ref. 118) is alsoshown. The WD andWescaleshavebeen arranged to
correspond to the same collision velocity. The measured cross section a for
electron transfer in H+-N2 collisionsis plotted vsWD = 2WH( )(ref. 117).
Thedestruction cross section equalsthe sumof the electron transfer
andionization crosssections (from ref. 116). convincing demonstration of this
notion wasmade by Koch,115'116 who examined the deexcitation, to n < 28,
and ionization or excitation ton > 61, offast D n = 46 atoms colliding with
N2. The 6-13 keV D2 beam energies gave the range of velocityratio6 < Vlv<
13.Thesumofthe excitation, ionization, and deexcitation cross sections is the
destruction cross section, which corresponds to the total e~-N2 scattering
cross section. As shown in Fig. 11.28, if we add the electron transfer cross
section for H+-N2 collisions,117to account for theRydbergioncore,

References 247 to the e~-N2 scattering cross section,118 the sum is in

excellent agreement with the Rydberg atom destruction cross section. In
particular, the N2~ resonance in the e~-N? scattering is reproduced well in
the Rydberg atom data. References 1. E. Fermi, Nuovo Cimento 11, 157
(1934). 2. A. Omont, /. Phys. (Paris) 38, 1343 (1977). 3. M. Matsuzawa, in
Rydberg States ofAtoms and Molecules, eds. R. F. Stebbings and F. B.
Dunning (Cambridge University Press, New York, 1983). 4. A. P. Hickman,
R. E. Olson, and J. Pascale, in Rydberg States ofAtoms and Molecules, eds.
R.F. Stebbings and F. B. Dunning (Cambridge University Press, New York,
1983). 5. M. R. Flannery, in Rydberg states of Atoms and Molecules, eds. R.
F. Stebbings and F. B. Dunning (Cambridge University Press, New York,
1983). 6. T. F. Gallagher, Rep. Prog. Phys. 51, 143 (1988). 7. T. F. O'Malley,
Phys.Rev. 130, 1020 (1963). 8. N. F. Lane, Rev. Mod. Phys.52, 29 (1980). 9.
M. Matsuzawa, /. Phys. B 13, 3201 (1980). 10. M. R. Flannery, Phys.Rev. A
22, 2408 (1980). 11. M. Matsuzawa, /. Chem. Phys.35, 2685(1971). 12. A. P.
Hickman, Phys.Rev. A 28, 111 (1983). 13. A. P. Hickman, Phys.Rev. A 19,
994 (1979). 14. F. Gounand, and L. Petitjean, Phys.Rev. A 30, 61 (1984). 15.
L. Y. Cheng and H. van Regemorter, /. Phys. B 14,4025 (1981). 16. A. M.
Arthurs and A. Dalgarno, Proc. R. Soc. London 256, 540 (1960). 17. A. P.
Hickman, Phys.Rev. A 18, 1339 (1978). 18. D. A. Park, Introduction
totheQuantum Theory (McGraw-Hill, New York, 1964). 19. M. Matsuzawa,
/. Phys. B 12,3743 (1979). 20. F. B. Dunning and R. F. Stebbings in Rydberg
States ofAtoms and Molecules, eds. R. F. Stebbings and F. B. Dunning
(Cambridge University Press, New York, 1983). 21. C. J. Latimer, /. Phys. B
10,1889 (1977). 22. E. de Prunele and J. Pascale,/. Phys.B 12,2511 (1979).
23. F. Gounand and J. Berlande, in Rydberg States ofAtoms and Molecules,
eds. R. F. Stebbings and F. B. Dunning (Cambridge University Press, New
York, 1983). 24. L. M. Humphrey, T. F. Gallagher, W. E. Cooke, and S. A.
Edelstein, Phys.Rev.A 18,1383 (1978). 25. T. F. Gallagher, S. A. Edelstein,
and R. M. Hill, Phys.Rev.A 15,1945 (1977). 26. F. Gounand, P. R. Fournier,
and J. Berlande, Phys.Rev. A 15,2212 (1977). 27. M. Hugon, P. R. Fournier,
and F. Gounand, /. Phys.B 12,1207(1979). 28. T. F. Gallagher, S. A.
Edelstein, and R. M. Hill, Phys. Rev. Lett.35, 644 (1975). 29. M. Harnafi
and B. Dubreuil, Phys.Rev. A 31, 1375 (1985). 30. R. Kachru, T. F.
Gallagher, F. Gounand, K. A. Safinya, and W. Sandner, Phys.Rev.A 27, 795
(1983). 31. M. Chapelet, J. Boulmer, J. C. Gauthier, and J. F. Delpech, /.
Phys.B 15, 3455 (1982). 32. M. Hugon, F. Gounand, P. R. Fournier and J.
Berlande, /. Phys.B 12,2707 (1979). 33. F. G. Kellert, K. A. Smith, R. D.
Rundel, F. B. Dunning and R. F. Stebbings,/. Chem. Phys. 72,6312(1980). 34.
C. Higgs, K. A. Smith, F. B. Dunning, and R. F. Stebbings,/. Chem.Phys.
75,745 (1981). 35. H. A. Bethe and E. A. Salpeter, Quantum Mechanics of
One and Two Electron Atoms (Academic Press, New York, 1957). 36. E. S.
Chang, Phys.Rev. A 31, 495 (1985).

248 Rydberg atoms 37. R. E. Olson, Phys.Rev. A 15, 631 (1977). 38. J. L.
Gersten, Phys.Rev. A 14,1354 (1976). 39. J. Derouard and M. Lombardi, /.
Phys. B 11,3875 (1978). 40. A. P. Hickman, Phys.Rev. A 18, 1339 (1978).
41. A. P. Hickman, Phys.Rev. A 23, 87 (1981). 42. T. F. Gallagher, W. E.
Cooke, and S. A. Edelstein, Phys.Rev. A 17,904 (1978). 43. M. P. Slusher,
C. Higgs, K. A. Smith, F. B. Dunning, and R. F. Stebbings, Phys.Rev. A 26,
1350(1982). 44. M. P. Slusher, C. Higgs, K. A. Smith, F. B. Dunning, and R.
F. Stebbings,/. Chem. Phys. 76, 5303 (1982). 45. T. F. Gallagher, R. E.
Olson, W. E. Cooke, S. A. Edelstein, and R. M. Hill, Phys.Rev. A 16, 441
(1977). 46. R. F. Stebbings, F. B. Dunning, and C. Higgs,/. Elec. Spectr. and
Rad. Phen. 23,333 (1981). 47. J. S. Deech, R. Luypaert, L. R. Pendrill, and
G. W. Series, /. Phys. B 10,L137(1977). 48. L. R. Pendrill, /. Phys.B
10,L469(1977). 49. A. C. Tarn, T. Yabuzaki, S. M. Curry, M. Hou, and W.
Happer, Phys. Rev. A 17,1862 (1978). 50. M. Hugon, F. Gounand, P. R.
Fournier, and J. Berlande, /. Phys.B 13,1585 (1980). 51. T. F. Gallagher and
W. E. Cooke, Phys.Rev. A 19,820 (1979). 52. T. F. Gallagher and W. E.
Cooke, Phys.Rev. A 19,2161 (1979). 53. J. P. Mclntire, G. B. McMillian, K.
A. Smith, F. B. Dunning and R. F. Stebbings, Phys. Rev. A 29, 381 (1984).
54. J. Boulmer, J. F. Delpech, J. C. Gauthier, and K. Safinya, /. Phys.B
14,4577 (1981). 55. M. Hugon, B. Sayer, P. R. Fournier, and F. Gounand, /.
Phys.B 15, 2391 (1982). 56. L. N. Goeller, G. B. McMillian, K. A. Smith,
and F. B. Dunning, Phys.Rev. A 30, 2756 (1984). 57. M. Hugon, F. Gounand,
and P. R. Fournier, /. Phys.B 13, L109(1980). 58. T. F. Gallagher, W. E.
Cooke, and S. A. Edelstein, Phys.Rev.A 17,125 (1977). 59. M. Czajkowski,
L. Krause, and G. M. Skardis, Can. J. Phys.51,1582 (1973). 60. E. Bauer, E.
R. Fisher, and F. R. Gilmore, /. Chem.Phys.51, 4173 (1969). 61. L. Petitjean,
F. Gounand, and P. R. Fournier, Phys.Rev. A 30,736 (1984). 62. L. Petitjean,
F. Gounand, and P. R. Fournier, Phys.Rev. A 30,71 (1984). 63. K. A. Smith,
F. G. Kellert, R. D. Rundel, F. B. Dunning and R. F. Stebbings, Phys. Rev.
Lett. 40, 1362(1978). 64. F. G. Kellert, K. A. Smith, R. D. Rundel, F. B.
Dunning, and R. F. Stebbings,/. Chem. Phys. 72,3179(1980). 65. A.
Kalamarides, L. N. Goeller, K. A. Smith, F. B. Dunning, M. Kimura, and N.
F. Lane, Phys.Rev. A 36, 3108 (1987). 66. H. Hotop and A. Niehaus, /.
Chem. Phys. 47, 2506 (1967). 67. C. A. Kocher and C. L. Shepard, /.
Chem.Phys.74, 379 (1981). 68. C. L. Shepard and C. A. Kocher, /.
Chem.Phys.78,6620 (1983). 69. R.D. Rundel (unpublished). 70. M.
Matsuzawa, /. Phys. B 55, 2685 (1971). 71. M. Matsuzawa and W. A.
Chupka, Chem.Phys.Lett. 50, 373 (1977). 72. T. F. Gallagher, G. A. Ruff, and
K. A. Safinya, Phys.Rev.A 22, 843 (1980). 73. G. Herzberg, Spectra of
Diatomic Molecules (Van Nostrand, New York, 1950). 74. A. H. Nielsen and
H. H. Nielsen, Phys.Rev. 48, 864 (1934). 75. A. H. Nielsen and H. H.
Nielsen, Phys.Rev. 14,118 (1938). 76. K. Rohr, /. Phys. B 13, 4897(1980).
77. W. P. West, G. W. Foltz, F. B. Dunning, C. J. Latimer, and R. F. Stebbings,
Phys.Rev. Lett. 36,854(1976). 78. B. G. Zollars, K. A. Smith, and F. B.
Dunning, /. Chem.Phys. 81, 3158 (1984). 79. G. F. Hildebrandt, F. G. Kellert,
F. B. Dunning, K. A. Smith, and R. F. Stebbings,/. Chem. Phys.68,1349
(1978). 80. L. Christophorou, D. L. McCorkle, and J. G. Carter, /.
Chem.Phys. 54, 253 (1971).

References 249 81. J. A. Stockdale, F. J. Davis, R. N. Compton, and C. E.

Klots, /. Chem. Phys. 60, 4279 (1974). 82. E. J. Davis, R. N. Compton, and
D. B. Nelson, /. Chem. Phys. 59, 2324 (1973). 83. W. T. Naff, C. D. Cooper,
and R. N. Compton, J. Chem. Phys. 49, 2784 (1968). 84. B. G. Zollars, C.
Higgs, F. Lu, C. W. Walter, L. G. Gray, K. A. Smith, F. B. Dunning, and R. F.
Stebbings, Phys. Rev. A 32, 3330 (1985). 85. G.W. Foltz, C.J. Latimer, G.F.
Hildebrandt, F.G. Kellert, K.A. Smith, W.P. West, F.B. Dunning, and R.F.
Stebbings, J. Chem. Phys. 67 1352 (1977). 86. C.E. Klots, Chem. Phys. Lett.
38, 61 (1976). 87. A. Chutjian and S. H. Alajajian, Phys. Rev. A 31,2885
(1985). 88. R.Y. Pai, L.G. Christophorou, and A.A. Christadoulides, J.
Chem. Phys. 70, 1169 (1979). 89. H. S. Carman, Jr., C. E. Klots, and R. N.
Compton, /. Chem. Phys. 90, 2580 (1989). 90. K. Harth, M.-W. Ruf, and H.
Hotop, Z. Phys. D 14, 149 (1989). 91. B. G. Zollars, C. W. Walter, F. Lu, C.
B. Johnson, K. A. Smith, and F. B. Dunning, /. Chem. Phys. 84, 5589 (1986).
92. I.M. Beterov, F.L. Vosilenko, I.I. Riabstev, B.M. Smirnov, and N.V.
Fateyev, Z. Phys. D 7, 55 (1987). 93. Z. Zheng, K. A. Smith, and F. B.
Dunning, J. Chem. Phys. 89, 6295 (1988). 94. C. W. Walter, K. A. Smith, and
F. B. Dunning, /. Chem. Phys. 90, 1652 (1989). 95. A. Kalamarides, C. W.
Walter, B. G. Lindsay, K. A. Smith, and F. B. Dunning, /. Chem. Phys.
91,4411(1989). 96. A. Kalamarides, R. W. Marawar, M. A. Durham, B. G.
Lindsay, K. A. Smith, and F. B. Dunning, J. Chem. Phys. 93, 4043 (1990). 97.
Y. T. Lee and B. H. Mahan, J. Chem. Phys. 42, 2893 (1965). 98. E. F.
Worden, J. A. Paisner, and J. G. Conway, Opt. Lett. 3,156 (1978). 99. J.
Weiner and J. Boulmer, /. Phys. B 19, 599 (1986). 100. J. Boulmer, R.
Bonanno, and J. Weiner, /. Phys. B 16, 3015 (1983). 101. A. N. Klucharev,
A. V. Lazavenko and V. Vujnovic, /. Phys. B 31143 (1980). 102. M. Cheret,
L. Barbier, W. Lindinger, and R. Deloche, /. Phys. B 15, 3463 (1982). 103.
K. Harth, H. Hotop, andM.-W. Ruf, inInternational Seminar on Highly
Excited States of Atoms and Molecules, Invited Papers, eds. S. S. Kano and
M. Matsuzawa (Chofu, Tokyo, 1986). 104. A. Mihajlov and R. Janev, /. Phys
B 14,1639 (1981). 105. R. K. Janev and A. A. Mihajlov, Phys. Rev. A
21,819 (1980). 106. B. G. Zollars, C. W. Walter, C. B. Johnson, K. A. Smith,
and F. B. Dunning, /. Chem. Phys. 85, 3132 (1986). 107. A. Kalamarides, C.
W. Walter, B. G. Zollars, K. A. Smith, and F. B. Dunning, /. Chem. Phys. 87,
4238 (1987). 108. L. Barbier and M. Cheret, /. Phys. B 20, 1229 (1987).
109. L. Barbier, A. Pesnelle, and M. Cheret, /. Phys. B 20, 1249 (1987). 110.
C. A. Kocher and A. J. Smith, Phys. Rev. Lett. 39, 1516 (1977). 111. J.
Boulmer, G. Baran, F. Devos, and J. F. Delpech, Phys. Rev. Lett. 44,1122
(1980). 112. D. Rapp and W. E. Francis, /. Chem. Phys. 37, 2631 (1962).
113. D. P. Hodgkinson and J. S. Briggs, /. Phys. B 9, 255 (1976). 114. R. D.
Rundel, D. E. Nitz, K. A. Smith, M. W. Geis, and R. F. Stebbings, Phys. Rev.
A 19, 33 (1979). 115. P. M. Koch, Phys. Rev. Lett. 41,99 (1978). 116. P. M.
Koch, in Rydberg States of Atoms and Molecules eds. R. F. Stebbings and F.
B. Dunning (Cambridge University Press, New York, 1983). 117. H. Tawara
and A. Russek, Rev. Mod. Phys. 45, 178 (1973). 118. R. E. Kennerly, Phys.
Rev. A 21, 1876 (1980).

12 Spectral line shifts and broadenings The first measurements of collision

properties of Rydberg atoms were the pressure shift measurements of Amaldi
and Segre,1 which prompted the description of the shifts in terms of free
electron scattering by Fermi.2 Lineshift and broadening measurements
provide information complementary to that obtained from the conventional
collision measurements described in the previous chapter. Just as the tunable
laser has made possible many of the collision experiments described in the
previous chapter, it has allowed much more sensitive line broadening
measurements. In this chapter we connect the normal description of
lineshapes to the Rydberg atom collision processes, briefly describe the two
modern experimental techniques, and describe the results of the experiments.
Theory If we consider the intensity of a transition from the ground state to a
Rydberg state, it has the form3 irrespective of whether it is a single photon
transition or a two photon transition. Our primary interest here is the impact
regime in which the Rydberg atom is colliding with one perturber at atime. In
this case the shift and broadening rates A and y are related to the shift and
broadening cross sections by3 y = 2N(Vah) (12.2a) and A = N(V<f), (12.2b)
where ob and as are the broadening and shift cross sections, V is the relative
collision velocity, and N is the perturber number density. We are particularly
interested in transitions from the ground state to the Rydberg states, in which
case the effects of collisions on the ground state are negligible, and only the
effects of collisions on the Rydberg states need be considered. First we
consider the broadening cross section ah. It arises from all collisions which
disrupt the atomic phase during the emission or absorption of 250

Theory 251 radiation. The more often the atomicphase is changed, the less
welldefined is the atomic frequency andthe broader the emission or
absorption line. Equivalently, the coherence between the upper and lower
levelsis destroyed.4 Its destruction is observed not only in line broadening
experiments, but in photon echo experiments aswell. For transitions between
the ground state and a Rydberg state of n > 10 there are three contributions to
ab. The first is inelastic collisions of the Rydberg electron with the perturber,
i.e. state changing collisions, which remove population from the Rydberg
state under study and halt the coherent absorption or emission of radiation.
Second, elastic collisions of the Rydberg electron withthe perturber
contribute to the broadening if the collisionally induced phase shift is large
enough, �jr. Third, collisions of the Rydberg ion with the perturber
contribute if the induced phase shift is large enough. Elastic electron-
perturber collisions obviously make no contributions to state changing
collisions, nor dothe Rydberg ion-perturber collisions make alarge
contribution, thus these cannot be observed in conventional depopulation
experiments, butonly in line broadening or echo measurements. We can write
thebroadening cross section explicitlyas5 where a^{is the Rydberg electron-
perturber elastic scattering broadening cross section, crinel is the
depopulation cross section due to the inelastic Rydberg electron-perturber
scattering, and ofon is the broadening cross section for Rydberg ion-
perturber scattering. We ignore the ground state contribution to the
broadening, and the factor of 1/2 is due to the fact that only the Rydberg state
contributes to thecross section. The elastic scattering cross section for a
perturber in state /? incident on a Rydberg atom in state awith momentum K
isobtained from the optical theorem. Explicitly3 Ajr obel =
�lm[f(a,l3,K^a,p,K)]- (12-4) We have already computed this cross section,
multiplied by the relative collision velocity V, inEq. (11.31). Explicitly
(12.5) where k is the momentum of the Rydberg electron. Using Eq. (11.14),
o{k) = 4jta2,where a isthe scattering length, and comes out of the integral of
Eq. (12.5), yielding l J |G(k)|2d^ W ^ . (12.6) The inelastic cross section is
simply the total depopulation cross section toall other states,i.e.

252 Rydberg atoms �inel = Odepop- (I2-7) Finally, the cross section for the
Rydberg ion-perturber scattering is obtained by computing the phase shift rj
due to the polarization interaction during the collision. Explicitly where
bisthe impact parameter of the collision between the Rydberg ion and the
perturber, ap is the polarizability of the perturber, and R is the separation
between the ionic core of the Rydberg atom and the perturber. Assuming a
straight line trajectory, the phase shift may be computed to be -^!i> (12.9)
and the cross sections for both the polarization broadening and shift can be
calculated using the method proposed by Anderson.6 Specifically,4'6'7 o\on
= 2JI r [1 - cosrj(b)]bdb, (12.10) Jo crfon = 2JZ sin r/(b)bdb (12.11) Jo
Using Eqs. (12.6), (12.7), and (12.10) wecan assemble the total broadening
cross section given by Eq. (12.3). The Anderson approximation bypasses the
ambiguous question of precisely how big a phase shift constitutes a line
broadening collision, or equivalently, precisely how the Weiskopf radius4 is
defined, removing an inherent ambiguity of the theory. The pressure shift
cross section has two terms, one from the elastic Rydberg electron-perturber
interaction and one from the ion-perturber interaction. Explicitly, a5 = a*, +
afon. (12.12) In some cases it is also convenient to break the shift itself into
two pieces, A = Aei + Aion = N(Va|,) + MVo?on>. (12.12a) We have already
computed asion in Eq. (12.11). In the line broadening literature the electron
contribution is usually described as3 = ^ Re [/(a,j8,K-> a,�,K)]. (12.13) We
can compute this cross section following the same approach used to compute
cjg!. Explicitly,

Theory 253 Re [/e(a,0,k-> aj8,k)] d3fc, (12.14) where // is the reduced mass
of the two colliding atoms and m is the mass of the electron. Using the fact
that Re[/e(a,/J,k)] = -a from Eq. (11.12), wefindthat Km or m Thisresult
wasfirstderived in an elegant manner by Fermi.2It can also be derived in a
pedestrian way using the form of the interaction given by Eq. (11.17). The
derivation is implicitly statistical, in that the motion of the perturber is
ignored, which at first seems to imply that it should not match the results of
an impact theory. However, we are only ignoring the thermal motion of the
perturber relative to the electron motion. It is still true that the Rydberg
electron interacts with one perturber at a time, thus the requirement of the
impact regime is met. The energy shift due to a rare gas atom at R due to the
electron-rare gas interaction is given by ' hm(r)UQ3, p,r,R)yjn�m(r) d3r.
(12.16) We can connect this energy shift per perturbing atom to an absolute
average shift by multiplying by the rare gasnumber density N and integrating
over the volume. If we also replace U(J},fi,r,R) using Eq. (11.19), Eq.
(12.16) becomes Aei = N ( yj*n�m(r)2jiad(r - R) VWm(r) d3rd3R = 2jzaN.
(12.17) In other words the shift is n independent and depends only on the rare
gas scattering length a and density N. It is often termed the Fermi shift. For
completeness, we also present the method used by Fermi2 to calculate the
shift due to the Rydberg ion polarization interaction in the statistical regime.
If there are rare gas atoms located at points R,-, the energy shift of a single
Rydberg ion is given by8 If there are N perturber atoms per unit volume, and
they are distributed uniformly, there will be one perturber per cube of
volume 1/N. If we assume that the perturbers are arranged in spherical shells
starting with radius Rx such that

254 Rydberg atoms we can convert the sum of Eq. (12.18) to an integral.
Explicitly, TM1 / 3 j . (12.20) This result,firstobtained by Fermi,2 has an
N4/3 dependence, markedly different from the linear density dependence
characteristically found in the impact regime. Experimental methods
Historically, the shift and broadening of spectral lines was thefirst way in
which collisions of Rydberg atoms were studied, by Amaldi and Segre.1
Much of the classical line broadening work hasbeen summarized byAllard
and Kielkopf,9 and we shall not describe it here. Using classical
spectroscopic techniques it was necessary tointroduce enough perturbing gas
that theshifts andbroadeningswere visible above the Doppler broadening, ~1
GHz. The development of Doppler free two photon spectroscopy makes it a
straightforward matter to observe lines which are lessthan 10 MHzwide,
reducing the required pressure bytwoorders of magnitude.10 The reduction in
required pressure ensures that measurements can be done in the impact
regime, not the quasi-static regime. In the latter case there are so many
perturbing atoms that simply considering their presence gives an adequate
representation of the observations. In the former casethe motion of the
perturbing atoms is important. Conventional lineshift and broadening
measurements are similar to absorption spectroscopy, in which the atoms
never need to be detected at all. All the information is inthe absorption of
thelight. If the atoms are detected, as isusually the case in two photon
spectroscopy, all that is necessary is discrimination between the ground state
and a Rydberg state, a trivial proposition. In many of the two photon,
Doppler free measurements the atoms are detected either by fluorescence
detection11 or, more often, using a thermionic diode12 in which collisions
convert the Rydberg atomstoions. As anexample,Fig. 12.1 depicts the
thermionic diode and single mode cw dye laser used by Weber and Niemax
to study self broadening in Cs.10Results typical of suchlinebroadening
experiments are shown in Fig. 12.2, in which the broadening and shift of the
two photon Rb 5s�> 24s transition with Ar pressure are quite evident.11 It
is also interesting to note that the shifts observed in thiswork would have
been marginally detectable, and the broadening unobservable, using
conventional absorption spectroscopy.
Measurements of shifts and broadening 255 Recorder Lock-in amplifier
Diode Confocal Fabry - Perot U 8 MHz Ar* Laser Dye Laser D::::: X
Optical isolator Oven Fig. 12.1 Thermionic diode and cw dyelaser used by
Weber andNiemaxtomeasure Cs self broadening (from ref. 10).
Thefinaltechnique whichhas been used extensively tomeasureline broadening
is the tri-level echo technique.12 The principle is shown in Fig. 12.3 as
applied to the Na ns1/2or nd3/2 states. At time tt aforward travelling laser
beam of frequency OJ1 excitesthe atomsfrom the 3s1/2state to the 3p1/2state,
and a beam propagating in the reverse direction, with frequency w2,
excitesthese atoms to an ns1/2 or nd3/2 state at time t2. This process forms a
coherent superposition of the ground state and the Rydberg state, and the
decay of thiscoherence isdominated by processes affecting the Rydberg state.
At a later time t3 the 3s-Rydberg coherence is transferred to a 3s-3p
coherence by a laser pulse of frequency a>2 propagating in the reverse
direction, and at time t4 � (t3 - t2)a)2/co1 + tx an intense echo pulse at co1
is emitted in the forward direction. The decay of the echo as the time interval
t3-t2isincreased reflects the decay of the 3s-Rydberg coherence. Echo
measurements are inherently Doppler free, and allthe information is
contained in the well directed optical echo beam. The atoms never need to be
detected at all. On the other hand the echo beam emerges from the sample
cell collinear with the o)x beam. For a sensitive detector to see the echo
pulse without being at least temporarily saturated by the w1 laser pulse
requires an electro optic shutter with an extinction of 107when closed.13
Measurements of shifts and broadening Extensive measurements of shifts and
broadening of transitions to alkali Rydberg statesbyrare gaseshavebeen
made. Arepresentative recent exampleis the use of

256 Rydberg atoms 1.96 Torr l J_ 0 100 200 300 400 500 600 MHz Fig.
12.2 Recorder trace of Rb 52s1/2 -> 242s1/2 signal at pressures of 0.29,
1.00, 1.45, and 1.96 Torr of argon, with different detector sensitivities, as a
function of laser detuning v. The sharp spikes are superimposed signals from
a 250MHz reference cavity (from ref. 11). twophoton, Doppler free
spectroscopy to measure both the shift and broadening of Rb 5s-�ns and 5s -
*nd transitions. In Fig. 12.4 we show the shift and broadening of Rb 5s �>
ns and 5s -� nd transitions by He observed by Weber and Niemax.14 In
Figs. 12.4 and 12.5 the shift coefficients are given in wave numbers per
number of atoms per cubic centimeter, or cm~Vcm~3 = cm2. As shown by
Fig. 12.4, the shifts of the transitions to both the ns and nd states are virtually
identical, and the shifts rise from zero at low n to a plateau at high n. Aside
from thefact that the shift at highn isindependent ofnand �,thefact that
itispositive is of historical interest. As we have already pointed out, the shift
has two components, the polarization shift and the Fermi shift.
Thepolarization shift is always negative, but the Fermi shift can be either
positive or negative depending on the sign of the electron scattering length of
the rare gasatom. When positive pressure shifts were observed by Amaldi
and Segre1 it was rather surprising as only the negative polarization shifts
had been anticipated. As shown by Fig. 12.4, the

Measurements of shifts and broadening 257 nd 3s time Fig. 12.3Tri-level

echoenergy leveland timingdiagram.The3satomsareexcited tothe nd Rydberg
state by two laser pulses of frequencies co1 and co2 at times tx and t^. The
a)x beam propagates intheforward direction (totheright) andtheco2
beampropagates inthe reverse direction (to the left). At time t3 a second
beam at w2 passes through the sample in the reverse direction, and an echo
beam at a)x emerges in the forward direction at Rb-He pressure shifts agree
fairly well with the pressure shifts predicted by Alekseev and Sobelman.3
Roughly 90% of the shift is due to the Fermi shift and 10% to thepolarization
shift. Thefractional values oftheseshifts aretypicalfor all the rare gases. The
shifts of the Rb transitions by He are in agreement with expectations. For Xe
though, shifts are somewhat different. As shown byFig. 12.5, the 5s �>
nsand 5s�> nd shifts both rise from zero and reach similar plateaus at n >
40.At high n, >40, the shifts match those calculated on the basis of elastic
scattering due to the polarization and Fermi interactions. However, the
5s�>ns shifts for n ~ 15 exceed the high n 5s �� ns and 5s�> nd shifts
and the 5s �> nd shifts for n ~ 15. Why are the n ~ 15 5s �> ns shifts larger
than the 5s �> nd shifts for n ~ 15, and why do they exceed the high n shifts,
which are usually the largest? A possible answer to both these questions
isthat the Rb nsstates of n = 20 are responding to the Xe not as separated,
noninteracting Rb+ and a Rydberg electron, but as a whole, easily perturbed
atom. The Rb nsstates have quantum defect 3.13 and lie just belowthe
degenerate high�states.Theytherefore aremuchmore susceptible
258 Rydberg atoms . -19 2, [10 cm 10 20 30 Fig. 12.4The measured
broadening and shift rates of the Rb n2s1/2 (O) and the ^dj levels (�) by He
asfunctions of the effective principal quantum number n*. Broadening: ( )
theory by Omont (ref. 16), ( ) Omont plus the theoretical inelastic
contribution by Alekseev and Sobelman (ref. 3) and ( - . - . - . - ) Omont plus
inelastic contributions measured byHugon etal. (ref. 15)(lower curve:�d,
upper curve:�s). Shift: ( )theoryby Alekseev and Sobelman (ref. 3) (from
ref. 14). toexternal perturbations, such asthepresence of aXeatom, than
arethendstates which have quantum defect 1.35. InFigs. 12.4 and 12.5 the
broadenings ofthe Rb5s �> ns and5s �> ndtransitions are also shown. If we
examine first the broadening by He, we can see that the broadening decreases
from the lowest states studied to a plateau at high n. The drop in the
broadening from n = 10 to n = 20 is due to the decline in elastic broadening
cross section, which is nearly the same for both ns and nd states with He. As
shown by Hugon et al. ,15the depopulation, or inelastic, cross sections for
Rb nsand nd collisions with He are negligibly small on the scale of Figs.
12.4 and 12.5. The high n plateau is due to the Rydberg ion-He scattering,
which is n

Measurements of shifts and broadening 259 N do'cm2] 15105Broodeninq i <\

\ Shift 10 20 30 40 50 10 20 30 40 50 Fig. 12.5 The measured broadening
and shift rates of the Rb w2s1/2 (o) and the ^dj levels (�) by Xe as
functions ofrc*.Broadening: (�) Omont's theory (ref. 16), ( ) Omont plus
inelastic contributions by Alekseev and Sobelman (ref. 3) and ( - . - . - . - )
Omont plus inelastic contributions measured by Hugon et al. (ref. 15) (lower
curve: /id, upper curve: ns). Shift:( ) theory by Alekseev and Sobelman (ref.
3) (from ref. 14). independent. The broadening is in reasonable agreement
with the theory of Omont,16 and the agreement isimproved
byaddingtheoretical3 or experimental15 corrections for inelastic collisions.
The broadening of the Rb 5s -> ns, nd transition by Xe is again different. For
both ns and nd states the broadening increases at low n, reflecting the
geometric cross section. The broadening peaks at n ~ 20for both ns and nd
states and then falls tothe sameplateau at highn, whichis dueto Rydberg ion-
Xe scattering. The major difference from He is that for n ~ 20 the broadening
is greater for the ns than the nd states. This difference is a reflection of the
larger inelastic Xe scattering cross sections of the Rb ns states. In fact, the
difference in the broadening cross sections of thens and ndstates from n = 32
ton =41 is the same asthe difference in the depopulation
crosssectionsmeasured by Hugon etal.15As shown by Fig. 12.5, the
agreement with the theoretical predictions is reasonable but not spectacular.
An interesting illustration of the importance of inelastic collisions and ion-
rare gasscattering is the broadening of theNans andndstatesobserved bythe
tri-level echo technique. In Fig. 12.6 we showthe rare gasbroadening of the
Na nsand nd

260 Rydberg atoms He � 3S-nD A 3S-nS (a) t 1000 CM o< b 600 z o o

#200 o cr o � r \ A � < A 4 i ' f A f A I 1 500 CNJ b 400 o �300 #200 o o
100 12 16 20 24 28 32 36 40 44 PRINCIPAL QUANTUM NUMBER n Ne
(b) * A * * A ? 4 8 12 16 20 24 28 32 36 40 44 n�*� Fig. 12.6 The
3s�>ns and 3s�>nd3/2 collisional broadening cross sections derived from
tri-level echo decay data areshown plotted vs theprincipal quantum number
of the upper state of the transition. Thecurved line corresponds to the
calculation of Omont (ref. 16). The flat part of the line is also calculated by
Alekseev and Sobelman (ref. 3).The errors shown aretypical andrepresent
thestatistical error in thedata (from ref.7).

CROS S SECTIO N <r(A 2) CROS S SECTIO N cr(& 2) � � 8 8 _ 00 ro

ro a. 00 Co ro ^ 3 ^0 . ^ ,_^_ , 3 O 3 rvi ^ ro ro f ' IN � I. r S I ON

262 Rydberg atoms Xe 2XI04 f CM o co in CO o o .X2 V.-V .� \ A) i f �

� :> � � k m \ (e) X � i i 4 8 12 16 20 24 28 32 36 40 44 n *> Fig. 12.6
Continued. states observed byKachru etal.7The broadening crosssections for
both nsand nd states rise at low n, reflecting the geometric cross section of
the Rydberg atom. After peaking, the ns and nd cross sections fall to the same
plateau at high n, reflecting Na+-rare gas scattering. The evident difference
between the ns and nd states is due to the fact that the nd states have large
inelastic � mixing cross sections,17"19 while the depopulation cross
sections of the ns states are nearly invisible on the scale of Fig. 12.6.20 If the
� mixing cross sections of the Na nd states are added tothe ns broadening
cross sectionsthe resulting cross sections are in reasonable agreement with
the nd broadening cross sections. An interesting experimental point is that the
�mixing crosssectionsgo tozero at highn whilethe broadening cross sections
reach a plateau due to the Na+-rare gas scattering. A graphic illustration of
the relative importance of the three components of the broadening
crosssections of the Na ndstates by He and Xeis found inthework of Gounand
et al.5 and shown in Fig. 12.7. In Fig. 12.7 are shown the calculated elastic
e~-rare gas, Na+-rare gas, and inelastic collision cross sections from
Hickman's scaling formulae.21 At low n the elastic cross sections can not
exceed the geometric cross section ~n4A2. From Fig. 12.7 is apparent that at
high n it is the Na+-rare gas interaction which dominates and at low n the e~-
rare gas interaction which dominates. If the samegraphsweretobe drawn for
the ns states the inelastic scattering cross sections would by several orders of
magnitude smaller since there are nearly no �mixing collisions of the Na
nsstates. A good illustration of the contribution of inelastic collisions to the
broadening crosssectionsisprovided by the alkaline earth atoms,
whichhaveperturbations in

Measurements of shifts and broadening 263 �< o 1O Fig. 12.7Relative

contributions to the cP values ofthe three terms ofEq. (12.3) inthe case of Na
nd states perturbed by He and Xe. Full line ( ) indicates the n independent
polarization term, dashed line ( ) theelasticpart duetothe (e~-perturber)
interaction, and dot-dashed line ( - . - . - . - ) shows theinelastic contribution
calculated according to the scaling formula of Hickman (ref. 21)(from ref. 5).
their quantum defects.22'23 In particular, in Sr the 5snd states increase their
quantum defects by 1 between n = 10 and n = 20.22 As a consequence, their
energy spacings from other 5sn� Rydberg states changes radically over this
range. Using two photon absorption spectroscopy Weber and Niemax22 have
measured the broadening of the Sr5s2 % -� 5sns % , 5s2% -> 5snd*D2, and
5s2 % -�5snd 3 D2, transitions by He and Xe. Where the 5snd quantum
defects match those of other Rydberg series increased broadening is
observed, due to the larger inelastic cross sections. A theoretical treatment,
based on quantum defect theory, of the variation in the broadening with n has
been presented by Sun etal.24 While pressure shift and broadening by rare
gases can be characterized as well understood, the same is not true of self
shift and broadening of the alkalis. The first two photon measurements of self
broadening of Rydberg states were made by Weber and Niemax,10 who
measured the self broadening of the Cs 6s1/2-� mi, 11 < n < 42 transitions
using a thermionic diode. Measuring typically every second or third n state
they observed an apparently smooth increase in the broadening rate up to
apeak of 13.8(33) MHz cm3at n = 25, then aslight decrease asn increases to
40. While the self broadening rate is an order of magnitude larger

264 Rydberg atoms , , 10 15 20 25 30 40 50 X CO Fig. 12.8 Observed

broadening and shift rates for the K 4s�� ns transitions by ground state K
atoms (from ref. 27). than for Xe, the most surprising aspect of the self shift
and broadening is the oscillatory structure in the alkali self shift and
broadening rates at n ~ 20, first observed clearly inRb byStoicheff
andWeinberger,25 who measured therates for every n. In retrospect, evidence
of the oscillations can be seen in the earlier measurements of Weber and
Niemax,26 but as they did not measure every n, the oscillations are not
apparent. The self shift and broadening of the Knsand ndstates provide a
good example of all the features mentioned above. Not surprisingly, the
broadening and shift rates are large, as shown by Figs. 12.8 and 12.9, plots
of the self shift and broadening of the K ns and nd states.27 At first glance the
shifts appear normal. They rise from approximately zero at low nton
independent plateaus at high n. However, the high n value is different for the
ns and nd states, an observation unlike anyofthe shifts observed withrare
gasesandinstark disagreement withthe usual theoretical description.
Furthermore, there is not a monotonic increase of theshift withn,rather
thereisapparent oscillatorystructure, asmentioned above. If we examine the
broadening of the levels, the gross characteristics of the nsand nd states are
the same and similar to rare gas broadening. The broadening of both ns and
ndstatesrises from zero atlown to amaximum atn ~ 20 and declinestothe same
value at high n. These characteristics are not different from those observed in
the Rb-Xe experiments shown in Fig. 12.5. The most striking feature of the
broadening is the pronounced oscillatory structure at n ~ 20. While the
structure is also present in the shift data, it is there not so pronounced.
Oscillations in the self shift and broadening similar to those

Measurements of shifts and broadening 265 10 15 20 25 30 35 7Fig.12.9

Observedbroadeningand shift ratesforthe K4s-^ ndtransitions byground state
K atoms(ref. 27). shown inFigs. 12.8 and 12.9 havebeen observed inK, Rb,
and Cs.25"28Previously Mazing and Serapinas29 had observed an
oscillation in the self broadening of the Cs6s-np transitions between n = 20
and n = 30. However, as shown byHeinke et al.,28 the origin of the
oscillation is the variation in proximity of a peak in the quasi-static line
wing. The oscillations shown in Figs. 12.8 and 12.9, obtained in the impact
regime, are not of the same origin. Matsuzawa30'31 has suggested that the
oscillatory structure may be due to narrow resonances in the e"-K scattering.
Asn is increased the lobes of the momentum distribution of the Rydberg
electron pass through the e~-K scattering resonances. In addition to
describing the locations of the oscillations, this explanation has the attraction
of implying oscillationsinboth the broadening and shift, with the
oscillationsinthe shift being about half as large. Kaulakys32 compared the
oscillations in the shift and broadening to those calculated on the basis of
theoretically predicted32'33 resonances in e~-K scattering. Kaulakys32 and
Thompson et al.,21 using the formulation of Matsuzawa, calculated
oscillations which matched those observed in the K self shift and broadening
experiments. The samestructure is observed when Rb is used to produce
thebroadening ofthe KRydberg states.27'28suggestingthatlow energy electron
scattering resonances in K~ and Rb" have the same energies and widths.

266 Rydberg atoms l6p,n/> I I6p,6p> ' 10" Fig. 12.10 Schematic diagram
showing the variation of the energy levels of asystem made of two Cs atoms
at distance xa0n2 from each other vs 1/x3 (from ref. 34). Herman34 has
suggested a more refined approach to this problem in which the correlation
between the Rydberg electron's position and momentum is included. Sincethe
negative ion resonance is apwaveresonance, herequires that there be a non-
zero derivative of the Rydberg atom spatialwavefunction wherethe resonant
electron momentum occurs. As n is increased from 15to 25 the orbital radius
at which the resonant momentum occurs moves inward from outside the
outermost lobe of the radial wavefunction, crossing several lobes of the
wavefunction and leading to the observed oscillations in the shift and
broadening. We have, until this point, discussed the broadening of the
transitions to Rydberg states by ground state rare gasand alkali atoms. Line
broadening due to other Rydberg atoms has also been detected by Raimond et
al.35and Allegrini et al.36In a Cs beam of ground state atom density 1012
cm"3, and a Rydberg atom density at line center of ~1010 cm"3 Raimond
etal.35have observed 30-40 GHz broadening of the transitions from the Cs
6p state to the Cs ns and nd Rydberg states of n ~ 40. Although the 6p �� ns
and 6p �> nd transitions are single photon transitions, the origin of the
broadening is due to two photon molecular transitions, an idea which is
easily appreciated by examining the molecular energy level diagram of Fig.
12.10. In Fig. 12.10, x isthe scaled interatomic distance, the
ratiooftheinteratomic spacingoftwo nearest neighbor atoms, Ry tothe size
ofthe Rydberg atom aQn2. The molecular states are composed of two Cs
atomic states. The low lying state 6p6p is unaffected by variations in x on the
scale shown, asis the molecular state 6pn�. However, the molecular state
n(n( and the nearby n'� � 1, n"� � 1state are significantly affected. These
doubly excited molecular states interact by means of the dipole-dipole
interaction,

References 267 tf�4� (12.21) where//' and//" are the electric dipole matrix
elements connecting the n� state to the n'� � 1 and n"(, � 1 states, andR is
theinternuclear separation. The twostates are shifted in energy by �Aa>.
When the strength of the dipole-dipole interaction of Eq. (12.21) becomes
larger than the R = o� energy spacingbetween the \n(,n() and |�'� � l,n"�
� 1) states the energy splitting between the two levels, 2Aa>, becomes
twice the dipole interaction of Eq. (12.21). If we make the approximation,
JU' = ju" = n2, and replace R by xn2 the splitting is given by 2Ao> = - 4 r
(12.22) x n If there were only these two states, their energy levels would
split with increasing 1/x3 as shown in Fig. 12.10. There are of course not
only two neighboring Rydberg states,but many, sothat inreality the entire
region spanned by the range 2Aco, contains the doubly excited nt,n'V states.
Thus 2Ao> should be a good estimate of the expected broadening. For a Cs
atom density of 1012 cm"3 and n = 40,x = 7,which using Eq. (12.22)
yields2Ao> = 3.6 x 10"6 = 24 GHz,in reasonable agreement with the
observed broadening. As shown byFig. 12.10the doubly excited n�,n�
states are, in general, excited from the 6p6p state by a two photon molecular
transition through a virtual intermediate state near the 6pn� state. However,
near the center of the 6p �> n�transition theymayalso beexcited via the real
6pn� state. It isinteresting tonote that thepressure broadening described here
isinessence the non-resonant manifestation of the same interaction
responsible for resonant Rydberg atom-Rydberg atom collisional energy
transfer.37 References 1. E. Amaldi and E. Segre, Nuovo Cimento 11,145
(1934). 2. E. Fermi, Nuovo Cimento 11, 157 (1934). 3. V. A. Alekseev and
1.1. Sobelman, Zh. Eksp. Teor. Fiz. 49, 1274(1965) [Sov. Phys.- JETP
22,882(1966)]. 4. 1.1. Sobelman, Introduction totheTheory ofAtomic Spectra
(Pergamon, New York, 1972). 5. F. Gounand, J. Szudy, M. Hugon, B. Sayer,
and P. R. Fournier, Phys.Rev. A 26,831 (1982). 6. P. W. Anderson, Phys.Rev.
76, 647 (1949). 7. R. Kachru, T. W. Mossberg, and S. R. Hartmann, Phys.
Rev.A 21, 1124 (1980). 8. H. Margenau and W. W. Watson, Rev. Mod.
Phys.8, 22 (1936). 9. N. Allard and J. Kielkopf, Rev. Mod. Phys.54,1103
(1982). 10. K. H. Weber and K. Niemax, Opt. Comm. 28,317 (1979). 11. W.
L. Brillet and A. Gallagher, Phys.Rev. A 22,1012 (1980). 12. T. Mossberg,
A. Flusherg, R. Kachru, and S. R. Hartmann, Phys.Rev. Lett. 39, 1523
(1977). 13. E. Y. Xu, F. Moshary, and S. R. Hartmann, /. Opt. Soc. Am. B 3,
497 (1986). 14. K. H. Weber and K. Niemax, Z. Phys. A 307,13 (1982). 15.
H. Hugon, B. Sayer, P. R. Fournier, and F. Gounand, /. Phys.B 15, 2391
(1982).

268 Rydberg atoms 16. A. Omont, J. Phys. (Paris) 38, 1343 (1977). 17. T. F.
Gallagher, S. A. Edelstein, and R. M. Hill, Phys.Rev.A 15, 1945 (1977). 18.
R. Kachru, T. F. Gallagher, F. Gounand, K. A. Safinya, and W. Sandner,
Phys.Rev.A 27, 795 (1983). 19. M. Chapelet, J. Boulmer, J. C. Gauthier, and
J. F. Delpech, /. Phys. B 15, 3455 (1982). 20. J. Boulmer, J.-F. Delpech, J.-
C. Gauthier, and K. Safinya, /. Phys.Rev. B 14,4577(1981). 21. A. P.
Hickman, Phys.Rev. A 23, 87 (1981). 22. K. H. Weber and K. Niemax, Z.
Phys. A 309, 19 (1982). 23. K. S. Bhatia, D. M. Bruce, and W. W. Duley,
Opt. Comm. 53, 302 (1985). 24. J.- Q. Sun, E. Matthias, K.- D. Heber, P. J.
West, and J. Gidde, Phys.Rev.A 43, 5956 (1991). 25. B. P. Stoicheff, and E.
Weinberger, Phys.Rev. Lett. 44, 733 (1980). 26. K. H. Weber and K. Niemax,
Opt. Comm. 31,52 (1979). 27. D. C. Thompson, E. Weinberger, G.-X. Xu,
and B. P. Stoichreff, Phys.Rev. A 35,690 (1987). 28. H. Heinke, J. Lawrenz,
K. Niemax, and K.-H. Weber, Z. Phys. A 312, 329 (1983). 29. M. Mazing
and P. D. Serapinas, Sov. Phys. JETP 33,294 (1971). 30. M. Matsuzawa, /.
Phys. B 10, 1543 (1977). 31. M. Matsuzawa, /. Phys.B 17,795(1984). 32. B.
Kaulakys, /. Phys. B 15, L719(1982). 33. A. L. Sinfailam and R. K. Nesbet,
Phys.Rev. A 7, 1987 (1973). 34. R. M. Herman, unpublished (1993). 35. J.
M. Raimond, G. Vitrant, and S. Haroche, /. Phys.B 14,L655, (1981). 36. M.
Allegrini, E. Arimondo, E. Menchi, C. E. Burkhardt, M. Ciocca, W. P.
Garver, S. Gozzini, and J. J. Leventhal, Phys.Rev. A 38, 3271 (1988). 37. K.
A. Safinya, J.-F. Delpech, F. Gounand, W. Sandner, and T. F. Gallagher,
Phys.Rev. Lett. 47,405(1981).
13 Charged particle collisions Due tothe longrangeinteraction between
aRydberg atom and achargedparticle, these collisions have by far the largest
cross sections of all nonresonant Rydberg atom collision
processes.Whilethere havebeen afewexperimentswithelectrons, most of the
the experimental work has been done with ions, because it ispossible to
make ion beams in which the ion is moving at a speed comparable to the
Rydberg electron's. Specifically A� and An state changing collisions,
ionization, and charge exchange have been carefully studied. State changing
collisions with ions State changing collisions between singly charged
positive ions and Rydberg atoms of Na have been studied extensively using
crossed beams of atoms and ions, the approach shown in Fig. 13.1.1 The Na
atoms in a thermal beam are excited to a HV PULSE eixy Fig. 13.1
Schematicdiagram of apparatus:IS,ionsource;O, atomicbeamoven; F, Faraday
cup; EM, electron multiplier; HW, hot-wire detector. Long and short dashed
line, ion beam; solid line, Na beam; dashed lines, laser beams (from ref. 1).
269

270 Rydberg atoms Rydberg state by two pulsed dye lasers tuned to the Na
3s �> 3p and 3p �> nd transitions. The excited Na atoms are bombarded
by an ion beam, which has an energy of 40-2000 eV, a current of 10-600 nA,
and a diameter of 5 mm. The Rydberg atoms are exposed to the ion beam for
times of 1-5 jus, after which the final states of the Na atoms are analyzed by
selective fieldionization. Since only low-� s,p, and dstatescanbe optically
excited, theinitial states areof low�andm and ionize adiabatically when
exposed to a field ionization pulse with a 1 JUS risetime. The final states are
most often the higher angular momentum states, which are composed mostly
of \m\ > 2 states, which ionize diabatically with the samefieldpulse. In all
experiments the ion beam axis has been perpendicular to the direction in
which the ionizing field was applied, so that the tendency of the collisions to
preserve m relative to the ion beam's direction still results in high m states
relative to the field ionization axis. In most experiments it has not been
possible to observe the difference in the adiabatic field ionization signals of
different �states of lowm, and collisions are typically detected bythe
presenceof a diabaticfieldionization signal. Na nd �> n( transitions The
most easily observed process with ionic, aswith neutral, collisions partnersis
�mixing. When the Na nd states are exposed to an ion beam,
thefieldionization signal changes from one which is predominantly adiabatic
to one which is predominantly diabatic. By measuring the fraction R of signal
transferred from the adiabatic to the diabatic peak of the field ionization
signal MacAdam etol. measured the depopulation cross section of the Na nd
states by He+ ions.1In the limit of small values of R the depopulation cross
section is given by1 o =^ (13.1) where/ isthe ion current density, Tis the
exposure time and e is the charge of the electron. InFig. 13.2we
showthevaluesofR, obtained with constant/ and T,as a function of n.
Equivalently, Fig. 13.2 is a plot of the cross section, in arbitrary units, vs n.
Since Fig. 13.2 is plotted on logarithmic scales, the cross sections evidently
exhibit a o <* n? dependence, and/? ~ 5. From the magnitudes of / and T, the
absolute value of the cross section for 450 eV He+ is determined to be 2.6 x
108 A2 for n = 28.The fact that the higher n cross sections at the ion energy
of 450 eV fall below the n5 dependence was later found to be an artifact due
to insufficient resolution of the diabatic and adiabatic field ionization
signals.2 In later experiments with other ions the n5 dependence shown in
Fig. 13.2 was also observed.2'3 The later measurements also verified that
the cross section was independent of ion species aslong as the ionshad the
samevelocity. Using ions of

� n� transitions 271 32 28 24 20 16 14 12 10 8 6 4 5 '. 600 eV y yl * *$ I

I/I /I 450 eV 28 24 20 R H 16 12 10 6 5 20 22 24 26 28 30 32 34
PRINCIPAL QUANTUM NUMBER Fig. 13.2 R, defined by Eq. (13.1)
andproportional toNand�-changingcross sections,vs n for 450 and 600
eVHe+ impact. Note thevertical offset of left and right logarithmicscales,
introduced for clarity. Different ion-beam currents were used at the two
energies. Leastsquaresfitsto the data are shown by the solid lines (from ref.
1). different mass the velocity dependence of the cross section can be
observed over a wide range of velocities, as shown in Fig. 13.3.2 Fig. 13.3
is a plot of the cross section for depopulation of the Na 28d state as a
function of the ratio of the ion velocity v to the Rydberg electron velocity, ve
= 1/n, in atomic units. Equivalently, ve = ca/n, and ve = 0.78 x 107 cm/s for n
= 28. As shown by Fig. 13.3, the cross section decreases as v/ve increases.
The observations are in good functional agreement with the theoretical
results of Percival and Richards,4 Herrick,5 and Shevelko etal.6 for v/ve >
1. Later calculations of Beigman and Syrkin7 agree with the experimental
cross sections to lower values of v/ve than do the theoretical results shown
in Fig. 13.3. The final state distributions, which yield additional insights into
the A�selection rules for the process, are equally as interesting as the cross
sections. When ion velocities of v ~ ve are used, for example 1000 eV Ar+
colliding with Na 28d atoms, there is a clear evolution of the diabatic field
ionization signal with increasing exposure of the atoms to the ion beam, as
shown in Fig. 13.4.8 At low current the diabatic field ionization signal has a
sharp peak, indicating that only a few states are populated, while at high
current the diabatic field ionization signal is quite broad, as would be
expected for all the n = 28, � > 2,\m\ > 2states. The fact that the field
ionization signal evolves with ion current indicates that the high current final
state distribution is due to multiple collisions. An analysis of the field

272 Rydberg atoms 20E o00 1 O C secti o cros s <D O) (0 V 16 12 8 4 l ~~

� i _ I I r \ 1 � H PR 1 1 H \ \ 1 \ SUV \ 1 | � Ar+ � Ne+ A He+ 1 1 0 1.0
2.0 3.0 4.0 Reduced velocity v/ve Fig. 13.3 Velocity dependence of the ^-
changing cross section. Error bars are the same for all points. Uniform 30nA
beams of He+, Ne+, and Ar+ at 400-2000 eV were incident on
Na28datoms.The datahavebeen normalized toPercival andRichard's (PR)
result (ref. 4) at v/ve = 1.5. The results of Herrick (H) (ref. 5) and Shevelko
etal. (SUV) (ref. 6) are also shown (from ref. 2). ionization curves shows
that the predominant initial transfer is into states of 3 < �< 5, leading to the
very narrow peak in the field ionization signals of Fig. 13.4 for low ion
currents.8 Measurements withNa+ ionsofenergiesin the29-590 eVrange,
corresponding to v/ve from 0.2 to 0.9, were compared to the diabatic SFI
spectra of the Na 28f, 28g, and 28h states observed individually by driving
resonant microwave transitions from the 28d state.9 These detailed
comparisons show clearly that for high velocity, v/ve^ 0.9, 59% of the 28d^
28� cross section is to the 28f state, the dipole allowed transition.
However, at lower values of v/ve, nondipole processes play a more
important role. For example, at v/ve = 0.2, only 37% of the cross section is
due to the 28d�>28f transition.9 At high velocities the process is
predominantly a dipole A� =1 process, but at low velocities the dipole
selection rule breaks down. Depopulation of the Na us and np states TheNans
andnpstates,havingquantum defects of 1.35 and0.85respectively, are well
removed from the hydrogenic �> 2states, and as a result the cross sections
a a Depopulation of the Na ns andnp states 273 I 1 1 1 1 1 1 I i i i r~ Ion
beam current (nA) 173 1000 1100 1200 1300 1400 1500 Electric field (V
cm"1) Fig. 13.4 Diabatic field ionization signals for � changing under
increasing incident beam intensities. For clarity the successive curves are
displaced upward by one scale unit. Data points are taken from the transient
digitizer records, and a small sloping background has been subtracted. Full
curves arefitsto the model of MacAdam et al.(from ref. 8). for collisionally
induced transitions from these states are smaller, by a factor of 10-30. The
collisions of ns states from 32s to 41s have been studied with both Ar+ and
Na+.10 At scaled velocities v/ve ~ 0.9 the cross sections exhibit an n4
scaling, but at the lowest values of v/ve studied the cross sections exhibit an
n5 scaling, as dotheNandstates.Thenpstates also exhibit cross sectionsscaling
as n4which are approximately one order of magnitude smaller than the nd
cross sections. Even so, these cross sections are about two orders of
magnitude larger than the geometric cross sections. One of the most
interesting aspects of the study of the Na ns and npstates is the distribution of
final states. In Fig. 13.5 we show the field ionization signals obtained when
the 39p,40s,39d, and 40pstates are exposed to 43 eVNa+ ions.10 There is an
initial adiabatic peak and a later broader diabatic feature. The Na+ current is
more than adequate to depopulate the 39d state, and the 39d signal
presumably reflects substantial population of thehigher �,m statesof n = 39,
due to both non-dipole low velocity collisions and multiple collisions. As
shown by

274 Rydberg atoms LU O 150 200 250 300 350 400 450 ELECTRIC FIELD
(V/cm) Fig. 13.5 Adiabatic and diabatic selective field ionization (SFI) for
^-changed ensembles produced from Na 39p, 40s,39d, and 40pstates by43
eVNa+ impact. The adiabatic peaks occur at 170-180 V/cm, and the diabatic
features occur above 250 V/cm (note change of vertical scale). The diabatic
SFI from �-changed 50s targets most closely resembles that from 39d. In
contrast, 40p and 39ptargets yields SFI that indicates a different distribution
of Stark sublevels lying high in the n = 39and 38 manifolds, respectively
(from ref. 10). Fig. 13.5, the 40s diabatic signal is approximately identical in
form to the 39d signal,indicatingthat nearly allofthen = 39
high�statesarepopulated from the s state. The extent to which multiple
collisions play a role is not so clear, but the final result isclearly similar to
that of the 39d state. In contrast, the np signals are very different. Much more
of the diabatic signal comes at higher fields. The fact that the s state and p
state signals are so different shows that the s state is not depopulated by
dipole transitions to the p states. Rather, collisional transfers are directly to
high � states. The fact that the p state adiabatic signal differs so markedly
from the d and ssignals also shows that the states populated from the p states
are not readily depopulated by subsequent collisions. This observation

Theoretical descriptions 275 suggests that these states might be the highest �
states, which have small dipole matrix elements connecting them to other
states. Theoretical descriptions At high velocities, v/ve > 1, these ion-atom
collisions can be described by the Born approximation, and this description
leads to areasonable description of the observed Nand�> ni results.4For
example, theresults of Percival and Richards,4 shown in Fig. 13.3, are
obtained by calculating the Born cross section for the nd-� ni transition and
using a sudden approximation to estimate the redistribution of population
among the nearly degenerate higher angular momentum states. At low
velocities, v/ve < 1, the Born approximation breaks down.Beigman and
Syrkin7have employed astrong coupling modelinwhichthe �and �� 1
states are coupled bythefieldfrom the passing ion. In itssimplest form their
model consists of three levels, the initial level �0, corresponding to the Na
nd state, and two degenerate higher �levels �0 + 1and �0 + 2.These three
states are coupled by the �� (� +1) electricdipolematrix elements
andthefieldofthepassingion. Instead of using the zero field�states as a basis
they use the three states7 K) Using this approach the |+) and |�) states are not
coupled bythefieldof the ion, but are only split in energy. At high collision
velocities the initial state |�0) is simply projected onto the |�0 + 1)state, a
coherent superposition of |+) and |�) states, by the dipole matrix element.
However, at lower velocities the change in energy of the |+) and |�) states
during the collision allows the |+) and |�) states themselves to be populated
rather than only acoherent superposition. The latter feature allows nondipole
transitions at lower collision velocities, as observed experimentally. The |+)
and |�) states of Beigman and Syrkin are Stark states, and their description
of slowcollisionsis related toonegivenoriginally by Smithetal? They
proposed that the collisional transfers occur if the field of the passing ion
brings the initial state to an avoided crossing with a Stark state of the
adjacent Stark manifold. This requirement is easily stated in a quantitative
form, % =ln2Eh (13.3)

276 Rydberg atoms where d{is the smallest absolute value of the quantum
defect of the initial state modulo 1; i.e. the energy separation of the initial /
state from the nearest high� states is dj/rc3. SinceE{ is given by 1/R2
whereR is the distance between the ion and the Rydberg atom, areasonable
estimate of the cross section isobtained byusing for R the impact parameter b.
Explicitly, o = nb2 = *-�-. (13.4) Although thissimple expression doesnot
allowfor anyvelocity dependence ofthe cross section, it does agree with the
observed zero field nd cross sections and implies that the Na nd cross
sections are a factor of 10larger than the np and ns cross sections, as
observed experimentally. Furthermore, this picture suggests that during the
collision the Rydberg high ( states are better described as Stark states than
�states, and that we might expect the (n + l)p and (n + l)s states to populate
the highest and lowest members of the n Stark manifold preferentially. In the
case of the s states there is no evidence to support this notion, but ion
collisions with the Na np states produce atoms which ionize with high
ionization fields, presumably due to diabatic ionization of either high
�states or higher lying Stark states. It is easy to believe that collisions
transfer the Na np state atoms to high Stark levelswhen R ~ b and that asthe
ion departs the Stark states relax to t states. In sum, this simple picture
allows us to describe many, but not all, of the features of state changing ion-
atom collisions. Electron loss The logical extension of the state changing
A� and Arc collisions is collisional ionization, for Na Rydberg atoms and
Ar+ the process Na n� + Ar+ -* Na+ +e~ + Ar+, (13.5) which isthe
dominant mechanism bywhich highvelocity, v> ve, ionsremove the electron
from aRydberg atom. On the other hand, when v/ve < 1 the incident ion
passesbythe Rydberg atom at a speed comparable to the Rydberg electron's,
and the electron is much more likely to attach itself to the incident ion than to
escape from both the Na+ and Ar+ ions.11 In other words, charge exchange,
the process Na nt + Ar+ -+ Na+ + Ar n't (13.6) is more likely to occur than
collisional ionization inslowcollisions.The sumof the cross sections for
ionization, aI? and charge exchange, acx, isusually termed the electron loss
cross section, aEL, i.e. Oi + acx. (13.7)
Electron loss 277 It corresponds to the crosssection for converting the
Rydberg atom to an ionwith no regard for the fate of the electron. At
highvelocities, v/ve � 1, the ionization crosssectionscanbe described using
the Born approximation, but the cross sections are small. In the region v~ ve,
where the cross sections are large, the Born approximation isnot valid, but
there are so many open quantum mechanical channels that rigorous coupled
channel calculations become impractical. Two theoretical approaches have
proven to be useful. One is to apply classical scaling arguments to existing
calculations of H+-H Is collisions. The essential notion is the following. If
the ratio of the velocity of the ion to the Rydberg electron's velocity, v/ve, is
the same as the velocity of the proton relative to the velocity of the H ground
state electron, then the ion-Rydberg atom cross section is n4 times as large as
the H+-H Is cross section, i.e. it is increased by the ratio of the geometric
cross sections. The second approach is to use classical trajectory Monte
Carlo techniques to calculate directly ion-Rydberg atom cross sections.11'12
In these calculations a set of initial conditions for the Rydberg electron is
chosen to correspond to the initial n and � of the Rydberg electron, and the
classical trajectory of the electron is computed as the ionpassesby.At the end
of the collision the electron is either still bound in the initial atom, free, or
bound to the incident ion, corresponding to no electron loss, ionization, and
charge exchange, respectively. Olson11 compared his Monte Carlo
Resultsfor n changingcollisionstothe analyticresults of Lodge et alP and
found excellent, 25%, agreement. In Fig. 13.6 we show the calculated charge
exchange and ionization cross sections for a Rydberg atom in states of n = 15
and �= 2 and 14 asfunctions of v/ve. Although the cross sections differ for
�=2 and 14,it is apparent that for v/ve < 1 charge exchange dominates,
while for v/ve > 1 ionization dominates.12Asshown by the scale ofFig.
13.6,thepeak cross sectionsfor both processes are approximately
equaltojcn4a02, the geometriccross section of the atom. The first electron
lossmeasurements were made byKoch and Bayfield14 using a fast beam
technique. They passed a H+ beam through a charge exchange cell of Xe to
produce collinear beams of H+ and neutral H atoms, in all states. By
modulating an axial field between 105and 171 V/cm they were able to
isolate, by field ionization, the signal from a band of about five n states
centered at n = 47. Prior to entering a high vacuum collision chamber, the H+
beam was decelerated to approximately the velocity of the H beam by a
second axial field. Accelerating plates after the collision chamber
accelerated the H+ ions. The H+ ions formed from H Rydberg atoms by
electron loss collisions had slightly different energies from those
oftheprimary H+ beam,which allowed themtobedetected separately from the
ionsin the far more intense primary H+ beam. Using this technique they
measured the electron losscrosssection for center of mass collision energies
from 0.2 to 60 eV, corresponding to v/ve from 0.3 to 3.3. The cross section
decreases from 8 x 10~9 cm2 to 1 x 10~9 cm2 over the velocity range
examined. For reference, the geometric cross section, Jtn4ao2, is 4.3 x 10"10
cm2. At high

278 Rydberg atoms 10 0.1 0.01 CHARGE EXCHANGE Fig. 13.6 Classical
trajectory Monte Carlo (CTMC) ionization and charge transfer cross
sections, with statistical standard deviations, for specific initial � plotted
against reduced impact speed v/ve. The cross sections are given inunits of
an2, where a^ = rra0. Circles are for � = 2, squares for � = 14. Also
included for comparison are points from approximate CTMC calculations
with the Na target core held fixed for � = 14 (triangles) and �= 2 (crosses)
(from ref. 12). velocities the observed cross sections have the same velocity
dependence as calculated cross sections15"17 but are a factor of 3.5 higher.
Measurements of electron loss from Na 40d and 30d states were done by
MacAdam et al.18 using crossed Na and ion beams, an arrangement similar
to the one shown inFig. 13.1. The resulting electron losscrosssections are
shownin Fig. 13.7, alongwith the H+-H n = 47 results of Bayfield and Koch
scaled by (40/47)4. The H+-H results exhibit the same v/ve dependence but
are a constant factor of about 3.5 larger than the other results. Asshown, the
crosssection for v/veis twice the geometriccrosssection. Severaltheoretical
crosssectionsfor electron loss and ionization are alsoshown inFig.
13.7.1519"22The theoretical ionization curves are the two which have cross
sections which increase with velocity at low

Charge exchange 279 Equivalent Energy for Stationary H(ls) Target

(keV/amu) _8 510 20 50 100 200 300 |x|0" 8 6 4 3 2 � IxlO"9 b 8 I l l (a\
Ad) / 1 1 1 o A (e) i 1 o ^g | 1 n = 40 o o T i i o ~~ 6 2 IxlO 0 1.0 2.0 3.0 4.0
Reduced Velocity v/ve Fig. 13.7 n =40 electron loss cross sections of
MacAdam et al.: solid circles (Xe+), triangles (Ar+), squares (Ne+).
Experimental results of ref. 13: open circles. Theoretical curves: (a) Janev
(ref. 19); (b) Percival and Richards (ref. 20); (c) CTMC (ref. 21); (d) Born
approximation (ionization only) (ref. 14); (e) 35-state close coupling
(ionization only) (ref. 22). Geometric cross section isog (from ref. 18).
velocities.15'22 The experimental and theoretical cross sections are in
excellent agreement for v/ve > 1. Only the theoretical cross section of
Janev19 goes to low velocities. It is calculated by applying classical scaling
laws to classical ionization and charge exchange cross sections. As shown
by Fig. 13.7, the theoretical cross section is somewhat larger at low
velocities than the experimental cross section. Wehave only discussed
electron loss produced by collisions of singlycharged ions withRydberg
atoms. Electron loss incollisionswithmultiply charged ions has also been
observed.23'24 but has not been studied extensively. Charge exchange As we
have already mentioned in the discussion of electron loss collisions and
shown inFig. 13.6,ionizing collisions of Rydberg atomswith lowvelocity ions
are

280 Rydberg atoms ION |0 N SOURCE OPTICS Fig. 13.8

Overallperspective viewof thechargeexchange apparatus.AthermalNabeam is
laser excited in the interaction region to produce a target for Rydberg-to-
Rydberg charge transfer experiments. The number of target Rydberg atoms is
determined by pulsed field ionization between the parallel plates and
bycollection of the resultant ionsinthe electron multiplier EM (from ref. 27).
most likely to result in charge exchange, and this process has been measured
directly by detecting the neutral product.25'26 The experimental arrangement
used to measure the final n state distribution subsequent to charge exchange is
shown in Fig. 13.8.27 A neutral beam of Na atomsis excited to an ns or
ndRydberg state bytwo pulsed dyelasers.An Ar+ ion beam, for example, with
an energy from 10 to 2000eV crosses the Rydberg atom target. Some of the
ions undergo charge exchange, resulting in the production of fast neutral Ar
n( Rydberg atoms which travel colinearly with, and at the same speed as, the
ion beam. The parent Ar+ beam is deflected from the fast neutral beam with
electrostatic deflector plates, and the fast neutral Ar Rydberg atom products
of charge exchange are detected byfieldionization. In measurements of total
charge exchange cross sections the stripper and 127� analyzer of Fig. 13.8
are replaced by a field ionizer with a 14 kV/cm field, one adequate to ionize
states as low as n = 15,ensuring ionization of virtually all the charge
exchange products from initial Rydberg states of n =25.26 The ions produced
by field ionization are accelerated to an energy of �14 keV, pass through a
thin C foil, and are detected by a particle multiplier. The C foil blocks
thepassage of slower (<1 keV) non-Rydberg neutral atomsresultingfrom
charge exchange with the background gas. The number of Rydberg atoms
initially produced by laser excitation is determined by field ionization of the
target Rydberg atoms, and the ion beam current is monitored with a Faraday
cup. Using these three signals, the geometry, and the timing of the experiment,
it is straightforward to extract the total charge exchange cross section. In Fig.
13.9 we show the relative charge exchange cross sections measured for Na+
ions impinging upon Na atoms initially in the 29s and 28d states, which are
nearly degenerate in energy.28 For v/ve > 0.8 the two cross sections
coincide, but for smaller v the 28d cross section falls below the 29s cross
section, except at the lowest velocity. Also shown are the classically scaled
cross sections for H+-H Is

Charge exchange 281 0) > (re CD X o D D.UU 2.00 1.00: 0.50 0.20
0.100.05i i i � I � � � V i i i i 1 t i � � 1 i N a Na I I � I I 1 � �
"I'+Na(29s) � " + + Na(28d) o �� P+H(1s)[9]� . \ ^ 4 I I � I V \ : 1 1 1
1 1 1 1 1 1 0.0 0.5 1.0 1.5 2.0 Reduced velocity v/ve Fig. 13.9 Velocity
dependence of charge transfer into all states. Na (28d) and Na (29s) targets.
The line represents experimental data of McClure (ref. 29) for H^ + H(ls)
electron capture scaled to match at v/ve = 1.0 (from ref. 28). charge
exchange,29 which are normalized to the experimental cross sections at v/ve
= 1. As can be seen in Fig. 13.9,the 29scross section agrees almost perfectly
with the scaled H ground state cross section, while the 28dcrosssection
obviously does not; it is smaller. Precisely why the 29s cross section agrees
with the scaled H+-H Is cross section and the 28d cross section does not
isunclear, but there are several possibilities. First, the 29s and 28d states do
not experience the same amount of collisional � mixing. Presumably the
initially populated 28d state has been spread over all then = 28, �> 2
states,whilethe29s state is not depopulated to the same extent. The 29s state
has zero angular momentum, as does the H Is state. In contrast, the � > 2
states have angular momentum and thus their orbits have classical outer
turning points at smaller radii than the 29s orbit, whichwould tend to make
the charge exchange cross section lower than the 29s cross section. Equally
as interesting as the total cross section is the distribution of the final Rydberg
state products of charge exchange. Since the cross section for v/ve � 1 is
approximately the geometric cross section of the target Rydberg atom, it is
not surprising that the final n state distribution is peaked near the n of the
target Rydberg atom.25 The experiments to measure final state distributions
are done using the apparatus of Fig. 13.8. The stripper and 127� analyzer
allow the analysis of the final Rydberg states. When an Ar+ ion beam is used
the parent Ar+ beam is deflected from the neutral Ar nt charge exchange
products by apair of deflecting plates, and onlythe neutrals enter the
concentriccylindrical stripper showninFig. 13.8. The outer cylinder is
grounded, and the inner concentric cylinder is held at

282 Rydberg atoms the voltage Vs, the stripper voltage. The field increases
as 1/r between the two cylinderswhere ris measured from the mutual axis of
the twocylinders. A neutral Ar n� atom entering through a hole in the outer
cylinder ionizes at the field corresponding to its nt state and is then
accelerated through a small hole in the inner cylinder. The field at which an
atom ionizes determines the energy it acquires in passing through the stripper.
The ions then pass through the 127� cylindrical energy analyzer set to pass
ions with kinetic energies 200eV less than eVs. Scanning the stripper voltage
over many laser shots yieldsthe distribution of fields at which the final states
ionize. Converting thefieldionization signals to an n distribution is a slightly
ambiguous process. In principle, charge exchange should result in no change
in m along the mutual axis of the ion beam and thefieldof the stripper, in
which casethe field ionization signals would be purely adiabatic if the initial
state was �< 2. With purely adiabatic ionization it would be possible to
make aunique conversion of a field ionization spectrum to an n distribution
using En ~ l/16rc4. However, the final state distribution is not inonly a few
low \m\ states. Thispoint wasverified by adding a 2.5 G magnetic field
perpendicular to the ion beam axis to cause the magnetic moments of the
atoms to precess about the fieldand alter the m values along the ion beam
axis. The presence of the magneticfielddid not alter the field ionization
spectra. Therefore we are forced to conclude that the final state distribution
is one consisting of a wide distribution of m states. This observation can be
reconciled with an expectation of a Am = 0 selection rule for charge
exchange by recalling that the deflecting fieldseparating the parent ions from
the product Rydberg atoms is transverse, allowing Am transitions. To convert
the observed field ionization spectra to n distributions two procedures were
employed.30 One, termed the "SFI centroid" approach, wassimply to
calculate the average field at which each n state is ionized, including both
adiabatic, \m\ ^ 2, and diabatic, \m\ > 2, contributions. This procedure
yields30 En = 3.3601 x 108n-38096. (13.8) Thefielddecreases lessrapidly
than n~4 because of the increasing number of high m states with increasing n.
This method allows a unique transformation from the field ionization
spectrum to an n distribution. A second method is to assume that thefinalstate
distribution is given by30'31 P(n) = Cf{n)e-^nVn~l)2 (13.9) where the
adjustable parameters n* and juspecify the peak and width of the distribution.
C is an arbitrary constant and/(ft) is a weighting factor. Choosing f(n) =
ncorresponds to allowing all�,but onlylowm,statestobepopulated, and
choosing f(n) = n2 corresponds to allowing all �ra states to be populated.
The maximum value of P(n) occurs at n = nmax. Uf(n) = 1, nmax = n*.
However, for the more physically likely cases/(�) = n and/(n) = n2, nmax >
n*. For example, for f(n) = n and n ~ 25 nmax ~ n* + 1. A synthetic
fieldionization signal is then computed using30

Charge exchange 283 Ar+(350 eV)+Na(28d) v/ve =0.526 stripper signal fit
� � 2 5 ~ 20 + > 15 + I 10+ \ 5+ o O 0 I n max=2 7 -2 SFI centroid fit � n
max=27'2 18 22 26 30 34 Principal Quantum Number n Fig. 13.10Charge
transferfinalstatendistributions following collisionsof350 eV Ar+ with Na
28d atomsobtained bythe SFIcentroidfitand the probability distribution of Eq.
(13.9) (fromref. 30). S(E) = XP{n)%{n,E), (13.10) where%(n,E) is the
probability that the state nionizes at thefieldE, assuming that the m < 2and m
> 2states ionize adiabatically and diabatically, respectively. The computed
values of S(E) are compared to field ionization spectra to determine P(n).
The difficulty is, of course, that the fields at which different n states ionize
overlap. Fortunately, using the two methods of analysis described above it
ispossible to obtain similar n distributions, as shown in Fig. 13.10. A very
interesting aspect of the fit using Eqs (13.9) and (13.10) is that/(rc) = n
produces a better fit than does f(n) = n2. This choice is consistent with the
electron in an initial low � Rydberg state's being captured into a state of any
� but low m. The distribution of the final states depends on the velocity of
the ion, as shown by Fig. 13.11. As v/ve is raised, the peak of the final state
distribution first rises then falls. In general, n* isvery near the n of the initial
Rydberg state. Also, n*is systematically higher for the (n + l)s state than the
nd state. Finally, the maximum value of ft* occurs for v/ve = 0.8. The widths
of the final state distributions also depend on v/ve. For example, of the three
distributions shownin Fig. 13.11, it is apparent that the one for v/ve = 0.751is
the narrowest. Finally, itis difficult to discern a significant difference
between the widths of the final state distributions for the (n + l)s and nd
states. The observed final state distributions are in reasonable agreement
with CTMC calculations of Becker and MacKellar,12 who calculated the
distributions for different initial values of �. In Fig. 13.12 we show the
theoretical final n distributions for target Rydberg atoms of n = 28 and
different values of � for

284 Rydberg atoms 35- 25 Ar+(350 eV)+ Na(25s) 3 0 "*" v/ve =0.444 n
max= 2 3 -5 /x=11.45�0.28 c 2 � � 10-H I i | I \ I i I | I ) i 1 I \ I i 35 (b)
Ar+(1OOO eV)+ Na(25s) 3 0 ^ v/ve=0.751 /x=16.40�0.40 �S 25-CD ^
20-b 15 + XJ c .2 10-18 22 26 30 34 38 42 46 50 54 Principal Quantum
Number n 22 26 30 34 38 42 46 50 54 Principal Quantum Number n 35 (c)
Ar+(21 00 eV)+ Na(25s) 3 0 T v/ve = 1.088 "�� 25-[- _rT~n-L n max=2 6
-2 /z=8.33�0.20 205 + 0 18 22 26 30 34 38 42 46 50 54 Principal Quantum
Number n Fig. 13.11 Final state histograms for the process, Ar+ + Na 25s,
demonstrating low, intermediate, and high energy behavior. The distributions
are obtained byusing Eq. (13.9) to fit the data: (a) 350 eV; (b) 1000 eV; (c)
2100 eV (from ref. 30).

Electron collisions 285 16 20 Fig. 13.12 Charge transfer cross sections

plotted against final n' at v/ve = 1 for the reaction Na+ + Na (n = 28,�)�>
Na (nf, all �') + Na+. The cross sections are expressed in terms of an2,
where an = n2a0. The distributions are shown for � = 2, 12, 27 and for an
initial distribution uniform in �2 (from ref. 12). v/ve = 1. As shown by Fig.
13.12, for � = 27 the theoretical final n state distribution is quite narrow,
while for � = 2 it is very broad. The distribution for all initial �states,
however, provides the best match tothe experimental data, suchas those of
Fig. 13.11. Since �mixingbythe ionshas a much larger crosssection than
charge exchange, most Rydberg atoms undergoing charge exchange are
equally likely to be in all �and m states. Although extensive charge
exchange measurements have only been done with singly charged ions,
charge exchange with multiply charged ions is now being explored.32
Electron collisions Electron collisionswith Rydberg atomsaresimilar
toioncollisionsin that they are of very long range. They differ, however, in
several important respects. First, charge exchange is clearly not possible,
leaving only state changing and ionizing collisions. As shown by Fig.
13.6,the ionization cross sections by ions peak when v/ve ~ 1. For electron
collisionsthe samecriterion applies, and for Rydberg atoms thermal, 0.01-
0.1eV, electrons are the ones which are most effective in inducing collisions.
Several indirect indications exist of the effects of electron collisions on
Rydberg atoms. Schiavone et al.33'34 produced Rydberg atoms by electron
bombardment of ground state atoms and observed atoms with high threshold
fields for ionization and such long radiative lifetimes that they must have
been

286 Rydberg atoms high m states of relatively low n. In addition, at very low
electron currents they observed anonlinear dependence ofthe Rydberg
atomproduction onthe electron current, suggesting that the long lived atoms
were being produced in two steps. First, low�Rydberg stateswereproduced
byelectron collisionswith ground state atoms, then in subsequent collisions
with electrons these atomswere collisionally transferred to high � states.
Since the cross sections for transitions between Rydberg states are large
compared to the cross sections for exciting the Rydberg states, only at very
low electron bombardment currents is the nonlinearity in the electron current
visible.They estimated the sizeofthe crosssection for �changing of n > 20
atoms to be 10~10-10~9 cm2.33'34 Similar observations of long lived,
presumably high �, states were reported by Kocher and Smith,35 who
excited Li atoms to Rydberg states by electron impact. However, they were
unable to determine if the high ( atoms were the result of single or multiple
electron collisions. Foltz et al.36 have made systematic measurements of
electron collisions with laser excited Na nd Ryderg atoms using an apparatus
similar to the one shown in Fig. 13.1 but differing intworespects.First,
anelectron beam isused instead ofan ion beam, and second, a magnetic shield
encloses the electron beam and interaction region. Rydberg Na nd atoms are
produced by two step pulsed laser excitation, via the 3p state. The atoms are
exposed to a 25 eV electron beam of current 1 juA and diameter 1cm for a
time of 6 JUS, after which the Na atoms are selectively field ionized. When
the atoms have been exposed to the electron beam, two new features appear
inthefieldionization signal,as shownbyFig. 13.13.The adiabatic feature at an
ionization field below the field of the initial nd state indicates the presence
of higher �,ra < 2 states, and the diabatic feature indicates the presence of
higher �, m > 2 states. Comparing these signals to the total Rydberg atom
signal allows the cross sections to be determined. The results are given in
Table 13.1 and compared to the theoretical results of Percival and Richards4
and Herrick,5 two of the theories shown in Fig. 13.3.As can be seen from
Table 13.1 and Fig. 13.3 both theories agree fairly well with both sets of
experimental data even though there is a difference of several orders of
magnitude in the cross sections. The difference inthe sizeof the
crosssectionsis duetothe relativelyhigh velocity of the 25 eV electrons. A 25
eV electron has a velocity of 3 x 108 cm/s, more than anorder of magnitude
higher than thetypicalionvelocities represented in Fig. 13.3.Since the cross
sections at high velocities decrease approximately as 1/v2 it is not surprising
that the observed electron cross sections are so much smaller than the ion
cross sections. While electrons in conventional beams have velocities too
high to have large cross sections, thermal electrons have large cross sections
for state changing collisions with Rydberg atoms, and these collisions have
been studied in a systematic fashion. Specifically, metastable He atoms in a
stationary afterglow have been excited to specific Rydberg states with a
laser.37'38 The populations of

Electron collisions 287 o 60 Co IO N unit ; < ^40 MO 3320 UJ u. 15 � 'o UJ

1 c � 5 UI u. y= Q (a) I1 �i f J \ � jl 1 Uhr In i \ m i i n = 50 ADIABATIC i
I iwllliL Tltfflit T %^ n = 50 DIABATIC 50 100 APPLIED ELECTRIC
FIELD (Vcrr Fig. 13.13 Typical selective field ionization data for laser
excited Na 50d atoms: (a) data with electron beam gated off (�), data
following collisions with 25 eV electrons (+) corrected for electron-induced
background signals; (b) net signal due to electron impact. The horizontal bars
indicate the range of field strengths over which n = 50 atoms are expected to
ionize adiabatically anddiabatically (from ref. 36). Table 13.1. Measured
and calculated cross sections for Na nd � changing collisions with 25 eV
electrons. Theoretical nd�> ni Experimental nd�> nia Percival &
Richards^ Herrickc n (cm2) (cm2) (cm2) 35 2.1 x 1(T9 2.0 x 10"9 36 1.6 x
10"10 40 1.5 x 10"9 3.6 x 10"9 3.5 x 10~9 45 3.0 x 10"9 5.8 x 10~9 5.7 x
10"9 50 3.4 x 10~9 9.0 x 10"9 8.9 x 10~9 a (from ref.36) *(from ref. 4) c
(from ref. 5)
288 Rydberg atoms all�statesof the samenwere monitored by time-resolved
np �> 2sfluorescenceto determine the total population and depopulation
rates by thermal electrons. From the results several conclusions maybe
drawn. First, the collisional mixing of the degenerate (m levels of the same n
is immeasurably rapid and istherefore at least two orders of magnitude faster
than An changing rates. From the buildup times of thefluorescencefrom n
levels other than the one populated by the laser the crosssections to those
levelscan be determined, and they are ~10~n cm2 and decrease rapidly with
increasing A/i. This dependence is in agreement with the Monte Carlo
calculations of Mansback and Keck39 and shows clear disagreements with
the more analytical approach of Johnson and Hinnov,40 which is based on
more restrictive assumptions, such as dipole transitions, and which predicts
a strong An = �1 selection rule. References 1. K. B. MacAdam, D. A.
Crosby, and R. Rolfes, Phys.Rev. Lett. 44, 980 (1980). 2. K. B. MacAdam,
R. Rolfes, and D. A. Crosby, Phys.Rev. A 24, 1286 (1981). 3. W. W. Smith,
P. Pillet, R. Kachru, N.H. Tran, and T.F. Gallagher, Abstracts, ICPEAC 13,
eds. J. Eichler, W. Fritsch, I. V. Hertel, N. Stotlerfoht, and U. Wille (North
Holland, Amsterdam, 1983). 4. I. C. Percival and D. R. Richards, /. Phys. B
10,1497 (1977). 5. D. R. Herrick, Mol. Phys.35,1211(1976). 6. V. P.
Shelvelko, A. M. Urnov, and A. V. Vinograd, /. Phys.B 9, 2859 (1976). 7. I.
L. Beigman and M. I. Syrkin, Sov. Phys. JETP 62, 226 (1986) [Zh.Eksp.
Teor. Fiz 89,400 (1985)]. 8. K. B. MacAdam, D. B. Smith, and R. G. Rolfes,
/. Phys.5.18, 441 (1985). 9. K. B. MacAdam, R. G. Rolfes, X. Sun, J.Singh,
W. L. Fuqua III, and D. B. Smith, Phys.Rev. ^36,4254(1987). 10. R. G.
Rolfes, D. B. Smith, and K. B. MacAdam, Phys.Rev.A 37, 2378 (1988). 11.
R. E. Olson, /. Phys. B 13, 483 (1980). 12. R. L. Becker and A. D.
MacKellar, /. Phys.B 17,3923 (1984). 13. J. G. Lodge, I. C. Pervival, and D.
Richards, J. Phys.B 9, 239 (1976). 14. P. M. Koch and J. A. Bayfield,
Phys.Rev. Lett.34, 448 (1975). 15. D. R. Bates and G. Griffing, Proc.
Phys.Soc. London 66, 961 (1953). 16. R. Abrines and I. C. Percival, Proc.
Phys.Soc. London 88,873 (1966). 17. D. Banks, PhD. thesis, University of
Stirling (1972). 18. K. B. MacAdam, N. L. S. Martin, D. B. Smith, R. G.
Rolfes, and D. Richards, Phys.Rev. A 34, 4661 (1986). 19. R. K. Janev,
Phys.Rev. A 28,1810 (1983). 20. I. C. Pervival and D. Richards,Adv.
Atomic and Molecular Physics 11,1(1975). 21. R. E. Olson, K. H. Berkner,
W. G. Graham, R. V. Pyle, A. E. Schlacter, and J. W. Stearns, Phys.Rev. Lett.
41, 163 (1976). 22. R. Shakeshaft, Phys. Rev. A 18, 1930 (1978). 23. H. J.
Kim and F. W. Meyer, Phys. Rev. Lett. 44,1047 (1980). 24. R. E. Olson,
Phys.Rev. A 23,3338 (1981). 25. K. B. MacAdam and R. G. Rolfes, /.
Phys.B 15, L243 (1982). 26. S. B. Hansen, L. G. Gray, E. Hordsal-Pederson,
and K. B. MacAdam, /. Phys.B 24, L315 (1991). 27. K. B. MacAdam and R.
G. Rolfes, Rev. Sci. Instr. 53, 592 (1982).

References 289 28. K. B. MacAdam, Nucl. Instr. and Methods B 56/57, 253
(1991). 29. G. W. McClure, Phys.Rev. 148,47(1966). 30. K. B. MacAdam,
L. G. Gray, and R. G. Rolfes, Phys.Rev. A 42,5269 (1990). 31. T. Aberg, A.
Blomberg, and K. B. MacAdam, /. Phys. B 20,4795 (1987). 32. B. D. De
Paola, J. J. Axmann, R. Parameswaran, D. H. Lee, T. J. M. Zouros, and P.
Richard, Nucl.Instr.and Methods B 40/41187 (1989). 33. J. A. Schiavone, D.
E. Donohue, D. R. Herrick, and R. S. Freund, Phys.Rev. A 16, 48 (1977). 34.
J. A. Schiavone, S. M. Tarr, and R. S. Freund, Phys.Rev.A 20,71 (1979). 35.
C. A. Kocher and A. J. Smith, Phys. Lett. 61A, 305 (1977). 36. G. W. Foltz,
E. J. Beiting, T. H. Jeys, K. A. Smith, F. B. Dunning, and R. F. Stebbings,
Phys. Rev. A 25, 187(1982). 37. J. F. Delpech, J Boulmer, and F. Devos,
Phys.Rev. Lett. 39,1400 (1977). 38. F. Devos, J. Boulmer, and J. F. Delpech,
J. Phys. (Paris) 40, 215 (1979). 39. P. Mansbach and J. C. Keck, Phys.Rev.
181, 275(1969). 40. L. C. Johnson and E. Hinnov, Phys.Rev. 181, 143
(1969).

14 Resonant Rydberg-Rydberg collisions Resonant energy transfer

collisions, those in which one atom or molecule transfers only internal
energy, asoppposed to translational energy, to itscollision partner require a
precise match of the energy intervals in the two collision partners. Because
of this energy specificity, resonant collisional energy transfer plays an
important role in many laser applications, the He-Ne and CO2 lasers being
perhaps the best known examples.1"4 It is interesting to imagine an
experiment inwhichwe can tune the energy of the excited state of atom
Bthrough the energy of the excited state of atom A, asshown in Fig. 14.1.5At
resonancewe would expect the cross section for collisionally transferring the
energy from an excited A atom to aground state Batom to increase sharply
asshown inFig. 14.1. In general, atomic and molecular energy levels
arefixed,and the situation of Fig. 14.1 is impossible to realize. Nonetheless
systematic studies of resonant energy transfer have been carried out by
altering the collision partner, showing the importance of resonance in
collisional energy transfer.6"8 The use of atomic Rydberg states, which have
series of closely spaced levels, presents a natural opportunity for the study of
resonant collisional energy transfer. One of the earliest experiments was the
observation of resonant rotational to electronic energy transfer from NH3to
Xe Rydberg atoms by Smith et al? Resonant electronic to vibrational energy
transfer has also been observed, from Rydberg states of Na to CH4 and
CD4.10 Both of these experiments, which are described in Chapter
11,correspond to tuning in steps equal to the discrete spacing of the Rydberg
levels. In an ideal study of resonant collisions it would be T B i Fig. 14.1
Twoatoms,A andB haveenergy levelsasshown. Initially atom A is
initsexcited state, and atom B is initsground state. If the energy of the excited
state of atom B couldbe tuned by some means,wewould expect the cross
section for resonant energy transfer from atom A to atom B to increase at
resonance as shown on the right (from ref. 5). 290

Resonant Rydberg-Rydberg collisions 291 478 479 800 Fig. 14.2 Energy
level diagram for the Na 16p, 17s, and 17pstates in a static electric field.
The vertical lines are drawn at the fourfieldswhere the s state is
midwaybetween the two p states and the resonant collisional transfer occurs
(from ref. 12). possible to tune continuously, as shown in Fig. 14.1, to
observe the collisional resonance explicitly. In fact, due to the large Stark
shifts of Rydberg atoms it is possible to tune them through a collisional
resonance. A good example is the resonant energy transfer in the collision of
two Na ns atoms in an electric field.n In Fig. 14.2 we show the energy levels
of the Na ns, (n � l)p, and np states as a function of electric field. In zero
field the ns state lies slightly above midway between the two p states.
However, as the field is increased the p states are shifted to higher energies
by their Stark shifts. As shown by Fig. 14.2, the field also lifts the
degeneracy of the \m\ = 0 and 1 levels of the p states. Here m is the azimuthal
orbital angular momentum quantum number. Due to the lifting of the m
degeneracy there are four fields at which the ns state lies halfway between
the two p states. At these fields two ns atoms can collide to produce an (n �
l)p and an np atom by the process Nans + N a n s ^ N a ( n - l)p + Nanp.
(14.1) As shown by Fig. 14.2, there are four collisional resonances for each
ns state. We label them by (m^, rau), where m� and rau are the \m\ values of
the lower and upper final p states. As shown by Fig. 14.2 in order of
increasing field the
292 Rydberg atoms np ns |b (a) (b) Fig. 14.3 (a)Energylevels for theNa ns, np
and (n � l)p statesshowing thedipole matrix elements coupling them, (b)
Diagram of the geometry of the collision in which atom2 passesbyatom 1
withvelocityv and impact parameter b(from ref. 5). resonances are (0,0),
(1,0), (0,1), and (1,1). The Na collision process described in Eq. (14.1) and
depicted in Fig. 14.2 is a resonant dipole-dipole process, and a simple
picture enables us to determine both the magnitude of the peak cross section
and thewidth ofthe collisionalresonance.Weconsider thecollision oftwo Na
atoms initially in the nsstate. During the collision they undergo transitions to
the np and (n � l)p states, by means of dipole transitions with the dipole
matrix elements JUX and ju2, as shown by Fig. 14.3. We assume that the
atoms follow undeflected straight line trajectories described by the impact
parameter b and collision velocity v, as shown in Fig. 14.3. For simplicity,
we first consider the special case of the staticfield'sbeing tuned to the
collisional resonance, i.e. to the fieldat which the energy of the ns state is
midway between those of the np and (n � l)p states. We wish to calculate the
probability of the atoms' making the transition to the np and (n � l)p states
during the collision. One approach is to treat one atom, atom 2, as providing
anoscillatingfieldwhich drivesthe transition in the other atom, atom 1.12In
other words, we treat atom 2as a classical dipole. Classically, adipole
transition matrix element becomes adipole of strength equal to the transition
matrix element oscillating at the transition frequency. Thus atom 2can be
thought of asacollection of dipoles corresponding to the ns �> np dipole
transitions allowed from the ns state. Most of these transitions are off
resonant when compared to the ns�>(n-l)p transition, and we ignore them.
The ns �> (n - l)p transition in atom 2, while degenerate can be ignored on
the basis of energy conservation (both atoms cannot make transitions to
lower energy states). Thus the only important transition of atom 2is the ns
�> np transition, in which case thefieldproduced by atom 2at atom 1 isgiven
by E2 = ju2/r3,ju2 being the ns-np dipole matrix element, r the distance
vector from atom 1 to atom 2, and

Resonant Rydberg-Rydberg collisions 293 r the internuclear separation

between the two atoms. This field is resonant with the ns �> (n � l)p
transition of atom 1 and drives the transition if the interaction is strong
enough. Since the dipole field falls quickly with increasing r, it is a
reasonable approximation to set V5 E2 =ju2/b3 r^ � b (14.2a) E2 = 0
r>�b. (14.2b) This choice of cutoff implies that thefieldfrom atom 2 is
present for atime x = b/v. In general atom 1undergoes the ns �� (n � l)p
transition if E 2 . ^ . r - 1 , (14.3) or if * & � - , - * � < i 4 - 4 > In Eqs.
(14.3) and (14.4) jux is the ns-(n � l)p dipole matrix element. Eq. (14.3)
makes it apparent that the requirement for driving the transition is that the
time integrated interaction be one. Solving Eq. (14.4) for b2, ignoring factors
ofJC, gives the cross section aR for the resonant collisional energy transfer.
Explicitly, (14.5) V Using the impact parameter b and the velocity v we can
also calculate the duration, or time, of the collision, (14.6) For the ns�> np
and ns �> (n � l)p transitions the dipole matrix elements are ~n2.
Accordingly, we insert n2 for jut and ju2 in Eqs. (14.5) and (14.6), finding
oR = nAlv (14.7) and x =n2/v3/2. (14.8) As shown by Eq. (14.7), the cross
section isgiven bythe geometric cross section divided by the collision
velocity. For thermal collisions v ~ 10~4, and as a result, aR~104rc4. For
n=20 this expression yields a cross section of ~10%)2, or ~2 x 108 A2,
which is large by atomic standards. Perhaps more interesting than the
magnitudes of the cross sections are the collision times. Evaluating Eq.
(14.8) for n = 20 and v = 10~4 gives r = 4x 108 or ~10~~9 s, a time far
longer than the typical atomic collision times of 10~12s. The discussion
above is focused on dipole-dipole collision processes, but it may easily be
extended to higher multipole processes. If atoms 1 and 2 have 2k pole
moments of n2k and n2k respectively, when the atoms are separated by
adistancer the interaction V due to these multipoles is

294 Rydberg atoms If we again assume that the atoms collidewith animpact
parameter band velocity v, following straight line trajectories , wecan again
assume that Vis only non zero for r ~ b and a time interval r = blv. Requiring
that Vr = 1 for a significant transition probability leads to For adipole-dipole
collision, k = k' = 1,and thisresult reduces to Eq. (14.5).On the other hand,
ask and k increase, the factor v2/("k+k'^ approaches 1 from below, andthe
crosssection decreases tothe geometricsize of the atom. It isworth noting that
the assumption of a straight line trajectory is not strictly valid in mixed
multipole collisions,orinanycollision inwhichtheinternal angular momentum
of the atoms is not conserved. However, the amount of angular momentum in
the translational motion is usually sufficiently high that small changes in it do
not significantly alter the trajectories of the colliding atoms. Two state theory
The expression of Eq. (14.7) for the cross section is adequate for a
calculation of the magnitude of the cross section, but it does not take into
account effects due to the orientation of the collision velocity v relative to the
tuningfieldE, nor doesit allow the calculation of the lineshape of the
collisional resonance. With these thoughts in mind, we present a more refined
treatment of the problem which enables us to describe some of the details of
the process. We shall consider the resonant collisions of two Na ns atoms to
yield an np and an (n � l)p atom, as indicated byEq. (14.1).The description
is,in principle,similar tothetreatmentof resonant rotational energy transfer
between polar molecules givenbyAnderson.8 It differs, however, in that the
presence of the electric field, which lowers the symmetry, must be taken into
account. Treatments specific to the problem of dipole-dipole collisions in
afield,such as the process depicted by Fig. 14.2,have been given by
Gallagher et al.,12 Fiordilino et al.,13 and Thomson et al.14 We assume
again that the atoms follow straight line trajectories, and we calculate the
transition probability, P(b), from the initial to the final state in a collision
with a given impact parameter, b. We then compute the cross section by
integrating over impact parameter, and, if necessary, angle of v relative to E
to obtain the cross section. The central problem is the calculation of the
transition probability P(b). The Schroedinger equation for this problem has
the Hamiltonian

Box interaction strengthapproximation 295 H=H0+V, (14.11) where Ho = Hi

+ H2 is the Hamiltonian for two non-interacting atoms in the static field, and
V is the dipole-dipole interaction v = Mi M2 r3 We construct the molecular
product stateswhichhave the spatial wavefunctions VB = Vlnp�1p2(n-l) P,
(14.13) where t/W andip2nt are theatomicwavefunctions of atoms 1 and 2 in
the n� state. These product states are the time independent solutions to Ho,
having energies WA = Wns + Wns and WB = Wnp + W(n_1)p. The dipole-
dipole interaction V has a diagonal term, which slightly shifts the energies,
andanoff diagonal term, which induces the collisional transitions. Since
Vdepends onr, Vis time dependent, and the wavefunction is given bythe
solution to the time dependent Schroedinger equation H\p = \d\pldt. (14.14)
When the two possible states are those given by Eq. (14.13) the general form
of the solution to Eq. (14.14) is V = CA{t)xpA + CB(0VB, (14.15) in which
all of the time dependence resides in the coefficients CA(t) and CB(t). Using
the wavefunction of Eq. (14.15) in the time dependent Schroedinger equation,
Eq. (14.14), we find two coupled equations for CA(i) and CB(t), WACA(t) +
VABCB(t) + VAACA(0 = iCA(0, (14.16a) WBCB(t) + VBACA(t) +
VBBCB(t) = iCB(0, (14.16b) where VBA = (VBI^I^A) = JVB VipA dr1 dr2
is the matrix element of V between the two spatial wavefunctions. In Eqs.
(14.16) the matrix elements of V are time dependent, since the internuclear
distance isafunction of time, asarethe coefficients CA(t) and CB(t). In
general, Eqs. (14.16) cannot be solved analytically, butin several special
cases analytic solutions can be obtained. Box interaction strength
approximation If we set VAA = VBB = 0 and use Eq. (14.2), i.e. assume that
VAB is a constant, = /a1//2/fo 3 , for r<V5b/2 and zero for r> V5Z>/2, as
given in Eq (14.2), the resulting approximation is equivalent to the usual
molecular beam magnetic resonance treatment.15 If we assume that r =
VV+v2*2 and that initially both

296 Rydberg atoms atoms are in the ns state so that ip(t0) = x/jA, CA(t0) = 1
and CB(t0) = 0 for t0 < 6/2v. After the collision, t > 6/2v, the probability
offindingthe atomsinthe p states, ip= I/>B> is given by ^ f f i (14.17) where
Ct = V(WA - WBf + A\VBA\2I2. Eq. (14.17) yields Lorentzian resonances
with linewidths �(Vv3//^/^)? m agreement with Eq. (14.6). While the
widths are meaningful, the lineshape, which depends on the form of the
interaction during the collision, is only approximate. At resonance WA =
WB, and Eq. (14.17) reduces to P(b) = sin2ffijgj, (14.18) in which we have
used the explicit form VAB = 3 Exact resonance approximation Another case
which may be treated analytically is the case of exact resonance, WA = WB,
if in addition, VAA = VBB = 0 and VAB = VBA, i.e. the coupling matrix
element is real. In this case Eqs. (14.16) are readily decoupled, leading to
two identical uncoupled equations. The equation for CA(t) is CA(t) = ^
CA(t) - V2ABCA(t). (14.19) The equation for CB(t) has identical
coefficients. If we again assume that initially, at t = -oo, CA(-oo) = 1 and
CB(-oo) = 0 the solutions to Eq. (14.19) which satisfy these boundary
conditions are12'13 ( ) (14.20a) and CB(O = siny VAB(t')dt'j. (14.20b) The
probability of making a transition is therefore given by VAB(t')dA- (14.21)
To obtain an explicit form for the matrix element VAB given in Eq. (14.12),
requires that wedefine the geometry of the collision. While the
collisionvelocityis normally the logical choice of quantization axis for field
free collisions, the
Exact resonance approximation 297 Z - y Fig. 14.4Diagram ofthe geometry
ofthe collision. The staticfieldis inthe z direction. The impact parameter
vector blies inthe x,z plane, at anangle 6from thez axis. The velocity vector v
liesinthe plane perpendicular to band makes anangle 0 with thex'
axis,whichis perpendicular tob andinthe x,z plane (from ref. 5). presence of
the tuning electric field makes the E field direction the logical choice, since
the eigenstates of H0 are easily described with this choice of quantization
axis. If we define the field direction as the z axis, so that E||z, then the impact
parameter vector b makes an angle 6with the z axis asshown in Fig. 14.4. We
may assume, with no loss of generality, that b lies in the x,z plane. The
collision velocity v, which is a constant vector due to the assumption of
straight line trajectories, is normal to b and makes an angle 0 with the x' axis,
which isinthe x,z plane and perpendicular to b, as shown in Fig. 14.4. If the
closest approach of the two atoms occurs at t = 0, the coordinates of atom 2
relative to atom 1 are givenby y = � vt sin 0 x = bsm6-vt cos 0 cos 6 (14.22)
z = b cos 0 + vt cos 0 sin#. Using these coordinate values we may now
evaluate the matrix elements of Eq. (14.12) by substituting for the dipole
moments the dipole matrix elements between the initial and final states. This
procedure yields explicitly time dependent matrix elements VAB(t). It is
particularly interesting to consider the (0,0) resonances, for two reasons.
First, the (0,0) resonances have no further splitting due to the spin orbit
interaction and are therefore good candidates for detailed experimental study.
Second, since these resonances only involve the matrix

298 Rydberg atoms elements ofjuz, which arereal,itis

possibletoevaluatethetransition probability at resonance analytically using
Eq. (14.21). The matrix element VAB (t)is given by T, / x /I 3Z>2cos2<9
6bvtcos 0cos 0 3v� cos20 sin20\ ^ A ^. VAB(t) = VZlVz2[^ p ^ ~ -f- 1 (14-
23) where r = Vb2 + v�\ In Fig. 14.5 we show graphs of VAB(t) for two
cases; v||E, and one example of vlE, for which 0 = 0 and 0 = JI/2. Note that
the specification v || E, specifies both 0 and 0. However vlE, only specifies 0
= JI/2; 0 can take any value. Theform of VAB(t) giveninEq. (14.23)canbe
integrated analytically tofindthe transition probability at resonance. Carrying
out the integration yields VAB(t')dtf =y^A (x _ 2 cos2 0 - cos20 sin2 0).
(14.24) vbz Evaluating Eq. (14.24) for v|| E, 0 =nil and 0 = 0, reveals that the
integral vanishes, as might havebeen anticipated from a closeexamination
ofFig. 14.5(a). The average value of VAB in Fig. 14.5(a) is clearly near zero.
On the other hand, for v�E, 0 = 0and 0 = jr/2,the integral of Eq. (14.24)
does not vanish, since the average value of VAB(t) doesnot vanish inthiscase,
as canbe seen inFig. 14.5(b). Correspondingly, at resonance, the cross
section must vanish for v || Ebut not for vlE. Numerical calculations
Normally one might expect that ifthetransition probability vanisheson
resonance it alsovanishes off resonance. However, such is not the case.
When the transition probability is calculated off resonance, by numerically
solvingEqs. (14.16)using a Taylor expansion method, it is nonzero for both v||
E and v1E.1416 In Fig. 14.6 we show the transition probabilities obtained
using two different approximations for v||E, and v l E for the 17s (0,0)
collisional resonance.16 To allow direct comparison to the analytic form of
Eq. (14.21) we show the transition probabilities calculated with VAA = VBB
= 0. For these calculations the parameters Vzi =/*z2= 156.4 ea0, b = 104a0,
and v = 1.6 x 10~4 au have been used. The resulting transition probability
curves are shown by the broken lines of Fig. 14.6. As shown by Fig. 14.6
these curves are symmetric about the resonance position. The v� E curve of
Fig. 14.6(b) has an approximately Lorentzian form, but the v|| E curve of Fig.
14.6(a), while it vanishes on resonance as predicted by Eq. (14.24), has an
unusual double peaked structure. Since the transition probabilities of Fig.
14.6 are obtained numerically, it is not appreciably more difficult to include
the diagonal matrix elements VAA and VBB arising from the permanent
dipole moments of the atomic states in thefield. The

Numerical calculations 299 -0.5 -10 -5 0 5 10 Time (b / v) -21 -5 0 5 10

Time (b/ v) Fig. 14.5Dipole-dipole interaction potential for (a) the case in
which v || E, (b) the casein which vlE with 0 = 0 and </> = nil (from ref. 14).

300 Rydberg atoms 0.0 -2.0 -1.0 Field From Resonance (V/cm) -2.0 -1.0
Field From Resonance (V/cm) Fig. 14.6Calculations of the transition
probability for the Na 17s + 17s -^ 17p + 16p, (0,0) transition for (a)v || E,
and (b)vlE. Solidlinesarefor calculationswhichincludethe effects of the
permanent dipole moments; dashed lines indicate that the permanent dipole
moments have been neglected (from ref. 16).

Intracollisional interference 301 solid lines of Fig. 14.6 are the transition
probabilities calculated with the permanent dipole moments at the field of the
collisional resonance taken into account. The permanent dipole moments of
the 16p, 17p, and 17s states are ju16p= 113.3ea0, ju17p = 182.3 ea0,
and//17s = �16.5 eoQ, respectively. Asshown by Fig. 14.6, when the
permanent moments are taken into account the locationof the resonance shifts
slightly, by approximately a resonance linewidth, asexpected from the fact
that the permanent moments are nearly as large as the transition moments.
Inaddition, the lineshapes become slightly asymmetric, due to the fact that the
energy spacing of the atomic states now changes with internuclear separation.
Intracollisional interference The unusual lineshape of Fig. 14.6(a) for v|| E
has approximately the same form and origin as theRamsey separated
oscillatory field pattern when there is a180� phase shift between the two
oscillatory fields.14 To see the similarity we first examine VAB(t), shown in
Fig 14.5(a). As shown by Fig. 14.5(a), for v|| E VAB changes sign twice
during the collision, sothat /"�, VAB(�')dr' = 0. Asign change of VAB(tf) is
equivalent to a 180� phase shift between the oscillatory fields ina separated
oscillatory field experiment, andthev|| E collisions arethus approximately
equivalent to doing an rf resonance experiment inwhich there are three
sequential rf fields with phase changes of 180�between successive fields.
At resonance, the contributions tothe transition amplitude from VAB(t') <
0cancel those from VAB(t') > 0so that, integrated over the whole collision,
the transition probability vanishes. Off resonance, however, the cancellation
is not complete. Calculation of crosssections To convert the transition
probabilities to cross sections wemust integrate over impact parameter using
a=\ P(b)2jvbdb. (14.25) Jo It is useful to consider the case of exact
resonance, which may be worked out analytically. At resonance, using either
Eq. (14.18) orEq. (14.21) the transition probability is given by P(b) =
sin21�), (14.26)

302 Rydberg atoms where D = pL^i2 if w e u s e Eq. (14.18) and D =

2juZljuZ2 (1 � 2 cos26 � cos20 sin2 0) if we use Eq. (14.21). In either
case D is proportional to the product of the two dipole moments. Examining
Eq. (14.26) we see that if we define b0 such that D/b02v =JT/2, as b
increases from 0 to b0, P{b) oscillates between 0 and 1, and as b increases
from b0 to infinity P(b) decreases smoothly to zero. If we insert the transition
probability of Eq. (14.26) into Eq. (14.25), the integral can be done
analytically, yielding 4 2 v Recalling that D ~ //i^2>w e see triat Eq. (14.27)
resembles Eq. (14.5) which was derived in a very simple way. If we are
interested inthe crosssection for v|| E, the angles 6and 0 are uniquely denned,
0 = jt/2 and 0 = 0. On the other hand, to calculate the observed v _L E cross
section Eq. (14.27) must be averaged over 0 for 0 = nil. In either case the
calculated cross sections must be averaged over the appropriate distribution
of collision velocities. In an analogousfashion thenumerically obtained
transition probabilitiesshown in Fig. 14.6 can be converted to cross sections
by integrating over impact parameter and averaging over the possible
collision velocities. Collisions with v1 Edo not have asingle allowed
valueof 0. However, sincethe lineshape ofFig. 14.6(b) issimple, some
averaging over the possible values of 6 has no appreciable effect. For v|| E, 6
and 0 are fixed, at 90� and 0�, so the integration over impact parameter is
only over ft, and the calculated v|| E cross section is very similar to Fig. 14.6.
The lineshapes of Figs. 14.6(a) and 14.6(b) are very different. However, this
difference is only apparent in an experiment with high resolution. At low
resolution the v || Eand v1 Ecollisional resonances can beexpected toexhibit
the same instrumentally determined shape. Furthermore, both cross sections
are approximately the same size. The calculated peak cross sections for the
Na 17s + 17s �> 16p + 17p collisional resonances are 6.0 x 10sa02 and 1.0
x 10%)2 for v|| E and v1 E respectively.14 When integrated over the tuning
field, to allow comparison to lowresolution experiments, the twointegrated
cross sections are 8.3 x 1O80O2 V/cm and 1.2 x 109a02 V/cm.14
Experimental approach Thebasicprinciple of the experimental approach is
easilyunderstood byconsideringthe Na collisions of Eq. (14.1) and Fig. 14.2
as a concrete example. As shown by Fig. 14.7,atomseffuse from aheated
oveninwhichtheNavaporpressure is �1 Torr.14 The Na atoms are collimated
into a beam by a collimator, not shown in

n scaling laws 303 Detector Atomic Source Electric Field Plates Laser
Beams Side View Fig. 14.7 Experimental apparatus configured with the
electricfieldperpendicular to the collisionalvelocity (from ref. 14). Fig. 14.7,
and pass into the interaction region, between a pair of electricfield plates, at
which point the beam density can be up to 109cm"3. The Na atoms are
excited to an ns state by two 5 ns pulsed dye lasers tuned to the Na 3s�> 3p
and 3p �> ns transitions at5890 A and �4150 Arespectively. Thens
atomsare allowed to collidewitheachother for aperiod of ~ 1JUS
duringwhichtimethe slowatoms in the beam are overtaken by the fast atoms.
Thus the collision velocity is approximately the width of the velocity
distribution of the beam. After the 1 JUS period during which the atoms
collide ahighvoltage pulse isapplied to the lower plate of the interaction
region. The amplitude of the pulse isset to ionize atoms in the np state, the
higher lyingfinalstate of the collision, but not atoms in the nsstate, the
initialstate of the collision. Ionsproduced byfieldionization areexpelled from
the region between the plates and impinge upon a particle multiplier above
the interaction region. The signal from the particle multiplier is recorded
with agated integrator. A static tuning voltage is applied to the lower plate to
tune the atomic levels into resonance. The collisional resonances are
observed by monitoring the field ionization signal from atoms in the np state
as the tuning voltage is slowly swept through the collisional resonances. A
typical example is shown inFig.14.8, a recording of thefieldionization
signalfrom the Na 17p statewhen the 17s state is populated by laser
excitation.12 n scaling laws One of the most striking aspects of resonant
collisional energy transfer between Rydberg atoms is the magnitude of the
cross sections. Accordingly, the first

304 Rydberg atoms 500 520 540 560 580 Fig. 14.8The observed Na 17p
ionsignalafter population ofthe 17s statevs dc electric field, showing the
sharp collisional resonances. The resonances are labeled by the |m| valuesof
the lower and upper p states (from ref. 12). experiments werefocused on
determining the magnitude of the crosssections and verifying the n scaling
lawsfor the Na ns + ns �> np + (n - l)p cross sections and collision times.
The cross sections were measured by taking advantage of the fact that the
number of atoms, iVp, which are in the npstate after the atoms have been
allowed to collide for a time interval T is given by12 where Ns is the
number of atoms in the ns state, Fis the volume of the sampleof excited
atoms, v is the average collision velocity, and a is the cross section. Eq.
(14.28) is valid only if thepopulation of the ns state is not depleted
significantly. If significant depletion of the population in the ns state occurs,
corrections to Eq. (14.28) must be introduced. The number of atomsinthe
npstate, Np, is related to the signal associated with np state, Np , by the
conversion efficiency of the detector, T. Specifically Np' = TNp, and Ws' =
TNS. Using Np' and Ns' we can rewrite Eq. (14.28) as " ^ \ 04.29) If the
geometry of the experiment, timing gates, oven temperature, and detector gain
are fixed, then the quantity in the square brackets of Eq. (14.29) is constant,
and the relative cross section as a function of n is easily obtained by
measuringthesignalsN'sandNp as afunction ofn. Therelative
crosssectionswere put on anabsolute basis bymeasuring the
quantitiesinthebracketsofEq. (14.29), with the result shown in Fig. 14.9. The
magnitudes and the n scaling of the

n scaling laws 305 ns 16 18 20 23 25 27 10 i i i i i i r~ 5.0 2.0 0.5 0.2 0.1 I I

I I I I 0.5 1.0 2.0 5.0 n*4(105) Fig. 14.9 The observed cross sections with
their relative error bars (�), and the fit curve (�)(fromref. 12). cross
section as n*37(5) are in good agreement with Eq. (14.7).12 Hererc*is the
effective quantum number of the ns state, defined by Wns = -l/2n*2. The
collisions resulting in the cross sections shown in Fig. 14.9 were most likely
not between atoms of different velocities in a directed beam but between
atoms moving inrandom directions. Both the excited atom sample volume and
collision velocity were probably a factor of 2 larger than thought at the time,
so the errors compensated to alarge extent. In any event, from Fig. 14.9it
isapparent that the cross section exhibits the expected n4 scaling and is of the
correct magnitude to match the expected a = n4/v behavior. Verification that
the collision time r increases asn2, or that the linewidth of the collisional
resonances decreases as n~2, is simply amatter of measuring the widths of
the collisional resonances. In Fig. 14.10 weshow aplot of thewidth of the
(0,0) resonance as a function of n. The observed widths exhibit magnitudes
and an n*-i.95(20) dependence in agreement with Eq. (14.8).

306 Rydberg atoms 0.2 0.5 1.0 2.0 n*4(105) Fig. 14.10The widths of the
(0,0) resonances (�), and the fit curve (�) (from ref. 12). Orientation
dependence of v and E and intracollisional interference One of the more
interesting aspects of the collision process is the calculated dependence on
the orientation of v relative to E. In particular, do the lineshapes for
collisions with v|| E and v1 E differ as dramatically as shown in Fig. 14.6?
To answer this question unambiguously required two improvements upon the
initial measurements. First, the Na atoms must be in a welldefined beam,
otherwise v is not well defined. This requirement is easily met by enclosing
the interaction region with a liquid N2 cooled box, which ensures that the
only Na atoms in the interaction region are those in the beam. Second,
thefieldhomogeneity must be adequate, ~1 part in 104, to resolve the intrinsic
lineshape of the collisional resonances. A pair of Cufieldplates 1.592(2)cm
apart with 1 mm diameter holes in the top plate to allow the ions to be
extracted is adequate to meet this requirement. Making these two
modifications, but keeping the basic geometry of Fig. 14.7, substantially
improves the resolution so that what appear to be four collisional resonances
in Fig. 14.8 are actually nine.14 All the resonances but the (0,0) resonance
are split by the spin orbit splitting of the np \m\ = 1 states. In the (0,1) and
(1,0) resonances the upper and lower p states, respectively, are split, leading
to doublets, while in the (1,1) resonance both upper and lower p states are
split, leading to a quartet. In view of the fact that only the (0,0) resonance is a
single line, it was selected for a detailed investigation of the cross section
with v|| E and v1 E. Tomeasure the crosssection with v1 Ethe geometry ofFig.
14.8,with the liquid N2 cooled shield, is adequate. To measure the cross
section with v|| E an arrangement, shown in Fig. 14.11, is used in which the
atomic beam passes

Velocity dependence of the collisional resonances 307 Atomic Source

ODetector Deflector Plates Laser Beams Top View Fig. 14.11 Experimental
apparatus configured so that the electricfieldand collisional velocity
areparallel (from ref. 14). through the field plates. In both cases the collision
velocity is the average of the difference in velocities of different atoms in the
beam. The observed crosssectionsfor the 18s (0,0)collisional resonance
withv || E and v 1 E are shown in Fig. 14.12. The approximately Lorentzian
shape for v1 Eand the double peaked shape for v|| E are quite evident. Given
the existence of two experimental effects, field inhomogeneties and collision
velocities not parallel to the field, both of which obscure the predicted zero
in the v|| E cross section, the observation of aclear dipin the center of the
observed v|| Ecrosssection supports the theoretical description of
intracollisional interference given earlier. It is also interesting to note that the
observed v|| E cross section of Fig. 14.12(a) is clearly asymmetric, in
agreement with the transition probability calculated with the permanent
electric dipole moments taken into account, as shown by Fig. 14.6. Velocity
dependence of the collisional resonances One of thepotentially most
interesting aspects of the resonant collisionsis that, in theory, the collision
time increases and the linewidth narrows as the collision velocity is
decreased. According to Eqs. (14.6) and (14.8) the collision time is
proportional to l/v3/2. Collisions between thermal atoms with temperatures
of ~500 K lead to linewidths of the collisional resonances that are a few
hundred MHz at n = 20. In principle, substantially smaller linewidths can be
observed if the collision velocity is reduced. The most straightforward way
of reducing the collision velocity is to velocity select the atoms in an
atomicbeam,17 and anatural method ofvelocity selection is

308 Rydberg atoms O C !? CO CO oo 368 369 370 371 372 373 374 375
Field (V/cm) o c CO CO oo 368 369 370 371 372 373 374 375 Field (V/cm)
Fig. 14.12 Experimental measurements of the Na 18s (0,0) resonance for (a)
v||E and (b) vlE. Note the definite dip near the center of the resonance in (a),
aswell as the slight asymmetry, both of which agree with the numerical
predictions (from ref. 14).

Oven Velocity dependence of the collisional resonances 309 Detector

Aperture Velocity selection wheel Field plates Fig. 14.13Experimental
arrangement for velocity selecting and focusing the atomic beam. The rotating
slotted disc and the pulsed laser beam select atomsinavelocity group, and the
hexapole magnet focuses them where they cross the laser beam (from ref.
18). the one shown in Fig. 14.13.18 Atoms effusing from the source pass
through a chopper, a rotating 9.6 cm diameter discwhich has radial slits as
narrow as 1.5 mm near the outer edge. The disc rotates at frequencies as high
as 200 Hz, at which frequency a 1.5 mm slit lets through a 25 jus wide pulse
of atoms. The laser, which has a 5 ns pulse width, excites the atoms 250
juslater at the interaction region, which is 10cm from the chopper. With a 1.5
mm wide slit the velocity distribution is ~10% of the velocity, which is
usually chosen to be the velocity of the peak of the velocity distribution of the
beam. Since a single velocity group of atoms is selected, it is possible to
focus them spatially using the hexapole focusing magnet, substantially
increasing the number density at the interaction region.1819 The first and
most obvious question is whether or not a narrower velocity distribution
leads to narrower collisional resonances. In Fig. 14.14we show the Na 26s +
Na 26s-^ Na 26p + Na 25p resonances obtained under three different
experimental conditions.20 In Fig. 14.14(a) the atoms are in a thermal 670 K
beam. In Figs. 14.14(b) and (c) the beam is velocity selected using the
approach shown in Fig. 14.13 to collision velocities of 7.5 X 103 and 3.8 X
103 cm/s, respectively. The dramatic reduction in the linewidths of the
collisional resonances is evident. The calculated linewidths are 400, 28, and
10 MHz, and the widths of the collisional resonances shown in Figs.
14.14(a)-(c) are 350,40, and 23 MHz respectively. The widths decrease
approximately as l/v3/2 until Fig. 14.14(c), at which point the
inhomogeneities of the electric field mask the intrinsic linewidth of the
collisional resonance. In the Na collision process of Eq. (14.1) the
inhomogeneities in the static tuning fields required preclude the observation
of very narrow collisional resonances. In

310 Rydberg atoms 46.0 48.0 50.0 52.0 Electric Field (V/cm) Fig. 14.14Na
26s + Na 26s�> Na 26p + Na 27p collisional resonances observed with (a)
no velocity selection, collision velocity 4.6 x l(r cm/s, (b) velocity selection
to a collision velocity 7.5 x 103 cm/s, (c) velocity selection to a collision
velocity 3.8 x 103 cm/s (from ref. 20). the K atom, however, dipole-dipole
resonances occur near zero electric field,18 where the homogeneity of the
staticfieldis less important. A K collision process which has been studied
extensively is171821 K29s + K27d-> K29p + K 28p, (14.30) which occurs
in relatively smallfieldsasshown in the energy level diagram of Fig.
14.15.Asshown by Fig. 14.15 andEq. (14.30), the29s and27d statesmustboth
be

Velocity dependence of the collisional resonances 311 -147.4 i � 29s, >,

-152.6 fe "153.7 i5 -158.8 27d, A 28P i/, I I I I I I 0.0 2.0 4.0 6.0 8.0 10.0
12.0 14.0 16.0 Electric Field (V/cm) Fig. 14.15 Energy leveldiagram for the
Rydberg statesof Kinvolvedin theresonant energy transfer
collisionsbeingstudied.Thetwo transitions shownhavebeenseparated for
clarity; they are degenerate in electricfield(from ref. 21). populated, and the
fact that a resonant collision has occurred is ascertained by detecting the
population in the higher lying29pstate byselectivefieldionization. As shown
by Fig. 14.15, the resonances occur near zero field, and it is easy to calculate
the small Stark shifts with an accuracy greater than the linewidths of the
collisional resonances. As a result it is straightforward to use the locations
of the collisional resonances to determine the zerofieldenergies of thep states
relative to the energies of thesand dstates. Sincethe energies of the ns and
ndstates have been measured by Doppler free, two photon spectroscopy,22
these resonant collision measurements for n = 27, 28, and 29 allow the same
precision to be transferred tothe npstates. Ifwe writethe quantum defect (5pof
the K np statesas <5P = d�p + dl{n*Y2 + (52(rt*)-4, (14.31) these
measurements have allowed a new determination of dp� with a factor of 5
smaller uncertainty than the previous value.23 Explicitly, they yield (5p�=
1.711925(3), in which the major uncertainly is from the Doppler free laser
spectroscopy, not from the resonant collision measurements.17

312 Rydberg atoms D C en C/) <N 6.4 6.5 6.6 6.7 Electric Field (V/cm) Fig.
14.16 Observed 29s resonance in K for several values of maximum-allowed
collision time: (a) r = 3.0 jus, FWHM = 1.4 MHz, (b), r = 2.0jus, FWHM =
1.8 MHz, (c) r = 1.0 ^s, FWHM = 2.2 MHz, (d), r = 0.4 jus, FWHM = 3.1
MHz, and (e) r = 0.2 jus, FWHM = 5.2 MHz (from ref. 21). Transform
limited collisions While collisional resonances 5 MHz wide are interesting
for spectroscopic purposes, what makes them most interesting is that the
5MHz linewidth implies that the collision lasts at least 200 ns, a time not
much less than the 1 JUS period allowed for the collisions to occur. If the
collision linewidths canbereduced to the inverse of the time allowed for the
collisions to occur, the collisional resonances become transform limited, and
we know when each collision begins and ends. Thomson et al. reached the
transform limit by incorporating two improvements.21 First, Helmholtz coils
were placed around the interaction region to cancel the earth's magnetic
field. Second, the time interval allowed for the collisions had previously
been defined by the laser pulse and the field ionization

References 313 pulse, which rises slowly at its onset. In these experiments
the interval was terminated by the application of a rapidly rising, ~50 ns,
low voltage detuning pulse which switches the field away from the field
value of the collisional resonance at awell defined time.24These
improvements allowed the observation of collisional resonances as narrow
as 1 MHz. These narrow resonances were then transform broadened by
reducing the time the atoms are allowed to collide to below 1 jus.In Fig.
14.16 we show collisional resonances observed as the time between the
laser pulse and the detuning pulse is shortened from 1 jus toOAjus, showing
clearly the transform broadening of the collisional resonances as the time
interval is shortened. If the collisional resonances are transform limited, we
know when individual collisions begin and end, and we can perturb the
colliding atoms atwell defined times during the collision, for example,when
the atoms are at their point of closest approach. References 1. C. K. N. Patel,
in Lasers, Vol. 2, ed. A. K. Levine (Marcel Dekker, New York, 1968). 2. A.
Javan, W. R. Bennet, Jr., and D. R. Herriot, Phys.Rev. Lett. 8, 470 (1962). 3.
A. D. White and J. D. Rigden, Proc. I.R.E. 50, 9167(1962). 4. C. K. N. Patel,
Phys. Rev. Lett. 13, 617 (1964). 5. T. F. Gallagher, Phys.Rept. 210, 319
(1992). 6. P. L. Houston, inAdvances in Chemical Physics, Vol. 47, eds. I.
Prigogine and S.A. Rice (Wiley, New York, 1981). 7. T. Oka, inAdvances in
Atomic and Molecular Physics, Vol. 9, eds. D. R. Bates and I. Esterman
(Academic, New York, 1973). 8. P. W. Anderson, Phys.Rev. 76, 647 (1949).
9. K. A. Smith, F. G. Kellert, R. D. Rundel, F. B. Dunning, and R. F.
Stebbings, Phys. Rev. Lett. 40, 1362 (1978). 10. T. F. Gallagher, G. A. Ruff,
and K. A. Safinya, Phys.Rev. A 22, 843 (1980). 11. K. A. Safinya, J. F.
Delpech, F. Gounand, W. Sandner, and T. F. Gallagher, Phys.Rev. Lett. 47,
405(1981). 12. T. F. Gallagher, K. A. Safinya, F. Gounand, J. F. Delpech, W.
Sandner, and R. Kachru, Phys. Rev. A 25, 1905(1982). 13. E. Fiordilino, G.
Ferrante, and B. M. Smirnov, Phys.Rev. A 35,3674 (1987). 14. D. S.
Thomson, R. C. Stoneman, and T. F. Gallagher, Phys.Rev.A 39,2914 (1989).
15. N. F. Ramsey, Molecular Beams (Oxford University Press, London,
1956). 16. D. S. Thomson, PhD Thesis, University of Virginia (1990). 17. R.
C. Stoneman, M. D. Adams, and T. F. Gallagher, Phys.Rev. Lett. 58, 1324
(1987). 18. M. J. Renn and T. F. Gallagher, Phys.Rev. Lett. 67, 2287 (1991).
19. A. Lemonick, F. M. Pipkin, and D. R. Hamilton, Rev. Sci.Instr. 26,1112
(1955). 20. M. J. Renn, R. Anderson, and Q. Sun, private communication
(1992). 21. D. S. Thomson, M. J. Renn, and T. F. Gallagher, Phys.Rev. Lett.
65,3273 (1990). 22. D. C. Thompson, M. S. O'Sullivan, B. P. Stoicheff, and
Gen-Xing Xu, Can. J. Phys.61, 949 (1983). 23. P. Risberg, Ark. Fys. 10, 583
(1956). 24. T. H. Jeys, K. A. Smith, F. B. Dunning, and R. F. Stebbings,
Phys.Rev.A 23, 3065(1981).

15 Radiative collisions A radiative collision is a resonant energy transfer

collision in which two atoms absorb or emit photons during the
collision.1Alternatively, aradiative collisionis the emission or absorption of
a photon from a transient molecule, and, as shown by Gallagher and
Holstein,2 radiative collisions can also be described in termsof line
broadening. In a line broadening experiment there are typically many atoms
and weak radiation fields, and in a radiative collision experiment there are
few atoms and intense radiation fields. The only real difference is whether
there are many atoms or many photons. Due to the short collision times,
~10~12 s, simply observing radiative collisions between low lying states
requires high optical powers, and entering the regime where the optical field
is no longer a minor perturbation seems unlikely. Due to their long collision
times and large dipole moments, Rydberg atoms provide the ideal system in
which to study radiative collisions in a quantitative fashion. As we shall see,
it is straightforward to enter the strong field regime in which the radiation
field, a microwave or rf field to be precise, is no longer a minor
perturbation. Ironically, while the experiments are radiative collision
experiments, with few atoms and many photons, the description of the
strongfieldregime isgiven in terms of dressed molecular states,which is more
similar to a line broadening description.3 Using a simple, three level model
we can develop a feeling for the microwave powers required to observe
radiatively assisted collisional energy transfer between Rydberg atoms.3
Consider the dipole-dipole atomic system shown in Fig. 15.1(a). In the Na ns
+ ns-^np + (n � l)p resonant collisions described in the previous chapter the
ns state corresponds to both s and s' of Fig. 15.1(a) and the n � 1 and np
states correspond to p and p' of Fig. 15.1(a), respectively. The collisions
occurs via the interaction V = ^ (15.1) In Fig. 15.1(b) we show the level
system of a collision radiatively assisted by the absorption of onephoton
offrequency OJ. Inthis casetheinteraction leadingtothe production of the p and
d states is given by where A is the detuningbetween thereal
andvirtualintermediate/?' states andEis 314

Radiative collisions 315 (a) s P1 (b) P' Fig. 15.1 (a) Energy levels and
dipole matrix elements for the resonant collision of two atomsinthe s and s'
statesresultingin theproduction oftwoatomsin thep andp' states, (b) Energy
levels and matrix elements for the radiative collision in which an s and an s'
atom collide to produce atoms in the p and d' states. The production of the d'
state is via the virtual p' state which is detuned from the real p' state by an
energyA. the amplitude of the microwave field in which the collision occurs.
The expressions of Eqs. (15.1) and (15.2) differ by the factor ju3E/k. For the
radiatively assisted collision shown in Fig. 15.1(b) to be ~10% as probable
as the resonant collision of Fig. 15.1a, we simply require that If we assume
typical values of ju3 � n2 and A ~ 0.1n~3, the required field is given by
Evaluating Eq. (15.4) at n = 20 leads to a field of 50 V/cm, or a power of ~6
W/cm2, a power roughly six orders of magnitude lower than those used in
optical radiative collision experiments with low lying atomic states.4"7
These powers are not only readily obtained, but are easily exceeded by
orders of magnitude, so that
316 Rydberg atoms 1 1 ~~ 18 p, mc = 0^�- 18 s - 17 p, mg = 0 < I I I y y �
LJ I �� I � I I I I l\ I I -369 -370 -371 -372 -373 -374 1 -395 > -396 en
-397 Lii -418 -419 -420 -421 -422 0 100 200 300 400 500 600 ELECTRIC
FIELD (V/cm) Fig. 15.2 Stark energy level diagram of the Na |m�| = 0
states, relevant to the multiphotonassisted collisions. The vertical lines
indicate the collisional transfer and are drawn at the fields where they occur.
The thick arrows correspond to the emitted photons (from ref. 8). itis a
straightforward matter toreach the strongfield regime,inwhichthe effect of the
field is no longer a minor perturbation.8 The Na ns + Na ns �> Na np + Na
(n - l)p collisions in combined static and microwave fields were the first
Rydberg atom radiative collisions studied. Specifically, the process Na 18s
+ Na 18s -> Na 18p + Na 17p + mco (15.5) has been examined in detail.
Here m is the number of photons emitted in the collision. When m > 0 there is
emission, andwhen m < 0 there isabsorption. The energy levels of the Na
18s, 17p, and 18p states are shown as a function of static field inFig.
15.2.8For simplicitywe haveonlyshownthe ra�= OlevelsinFig. 15.2. The
(0,0) resonant collision, with no microwaves, occurs at a field of 390 V/cm,
and the radiative collisions accompanied by the stimulated emission of one,
two, and three photons occur at lower staticfields, as shown.Wefollow the
convention of labelling the collisional resonancesby them�valuesofthelower
andupper final pstates of the collision, i.e. by (mhmu), where mxand rau are
the m� values of the lower and upper/? states. The number of photons
emitted is shown by the number of short bold arrows. In Fig. 15.2 the field
displacement of a radiative collision resonance from the normal collisional
resonance is proportional tothe microwave frequency and inversely
proportional to the sum of the differential Stark shifts of the/? states. As
shown inFig. 15.2, the 18s state has avery small Stark shift, which weignore.
In principle, it is equally possible to observe collisionsinwhich there is
absorption, as opposed to emission, of several microwave photons.
However, in

Initial experimental study of radiative collisions 317 the Na ns + Na ns �>

Na np 4-Na (n � l)p system these radiatively assisted collisions occur at
higher staticfields,where there are manycollisional resonances due to the
Stark states which are not shown in Fig. 15.2.9The n = 17 and n=16 Stark
manifolds lie just below the 18p and 17p states, and at fields�10%higher
than the fields at which the resonant collisions of Fig. 15.2 occur, resonant
collisions involving the Stark states appear. The presence of these collisional
resonances precludes the unambiguous observation of ns + ns �> np + (n �
l)p collisional resonances in which microwave photons are absorbed. Initial
experimental study of radiative collisions The initial experimental study of
radiative collisions between Rydberg atoms was done by simply introducing
a microwave field into the region between the two field plates of an
apparatus such as the one shown in Fig. 14.7.10 The microwave field was
introduced using a horn outside the plates, and several microwave
frequencies between 12 and 18 GHz which propagated well between the
plates were used. The maximum intensities were estimated to be ~3 W/cm2.
The data are accumulated in the same fashion as in the resonant collision
experiment, the only difference being the presence of amicrowavefield.The
atoms are excited by twopulsed lasers, and collisions are allowed to occur
for a 1^s period, after which ahighvoltage pulseis applied tothe
lowerfieldplate tofieldionize atomsinthenp state. The signal from the np state
atoms isrecorded asthe staticfieldis scanned. On each scan the
microwavefieldamplitude isheld constant. The dependence of the radiative
collision cross section on the microwave power is determined by repeating
the staticfieldscan atdifferent microwavepowers. Inthese experiments single
photon radiative collisions were observed, at staticfieldsslightly below the
fields at which the resonant collisions were observed, as expected from Fig.
15.2. A typical example is shown in Fig. 15.3. The field displacement of the
radiative collision signals from the resonant collision signal isequal to
thefieldrequired to produce a Stark shift equal to the microwave frequency.
Two points about this experiment areworth noting. First, the radiatively
assisted collision crosssections are within an order of magnitude of the
resonant collision cross sections for incident microwave intensities of a few
W/cm2, confirming the estimate of Eq. (15.4). Second, it is interesting to
observe that at low microwave power the radiatively assisted collision
signal islinear in the microwave power, but at higher powers the radiatively
assisted collision signal rises less rapidly, suggesting the entrance into a
nonperturbative regime. Replacingthefieldplates ofFig. 14.7withtheresonant
microwave cavityshown in Fig. 15.4 allowed an increase in the circulating
microwave power by Q, the quality factor, of the cavity.8 The cavity is
apiece of WR-90 (X band) waveguide 20 cm long which is closed at both
ends. The inside dimensions of the cavity are
318 Rydberg atoms AW(crrT1) < zSIG z o N Z o 0.4 1 22 s (a) 15 GHz 9 W
J*l ilk, AA M(:�*{...�..\..,J� I 0.2 I 0.0 I 1 A 1 -0.2 1 A \in -0.4 1 1 100
120 140 DC FIELD (V/cm) Fig. 15.3 The observed Na 22pion signal after
the population of 22sstate vsthe dcfieldin thepresence of a 15
GHzmicrowavefield(solidtrace).The dotted tracewas observedwith no
microwave field, and the sharp resonances in the center (for both solid and
dotted traces) result from the resonant collisions,whilethe displaced
resonances onthe side (solid trace) are due to microwave assisted collisions.
The scaleonthe top of thefigureshowsthe detuning from the (0,0) resonance
(from ref. 10). Signal Out Fig. 15.4 Main features of the atomic beam
apparatus, theatomic source, the microwave cavity, and the electron
multiplier. The microwave cavity is shown sliced inhalf. The Cu septum
bisects theheight of the cavity. Two holes of diameter 1.3 mm aredrilled in
the sidewallsto admitthecollinear laser andNaatomicbeams,anda1
mmdiameter hole inthe top ofthe cavity allowsNa+ resulting
fromfieldionization of Na to be extracted. Note the slots forpumping (from
ref. 8). 20.32 cm long, 1.02 cm high, and 2.28 cm wide. The cavity contains a
septum, as shown inFig. 15.4, whichallowsthe application of a static
tuningfieldand a pulsed field inthe vertical direction. Thepresence of the
septum doesnot affect the TE1On cavity modes since they have only vertical
electric fields. The cavity is operated on modes of odd n, which have an
electricfieldantinode at the center of the cavity. Specifically the n = 15,17,
and 19modes,with resonant

Initial experimental study of radiative collisions 319 frequencies of 12.8,

14.4, and 15.5 GHz have been used. These modes have Q ~ 1300, and when
excited with a power of 20 W, the circulating microwave intensity is ~10
kW/cm2, and the microwavefieldis ~500 V/cm. Asshown inFig. 15.4, the
atomic and laser beams enter the cavity through opposing holes in the
sidewalls of the cavity ~1 mm above the septum. Thus a cylindrical volume
of Rydberg atoms is produced. There isasmall, ~1 mm diameter, hole in the
center of the top of the cavity through which ions produced byfieldionization
of atoms directlybeneath theholeescapefrom the cavityen route tothe detector.
Since the atoms only travel ~1 mm during the time between laser excitation
and field ionization, the location of the hole inthe top ensures that the
signalsobserved are due to atoms at an antinode of the microwave electric
field. Data are accumulated in the same way as used to acquire those shown
in Fig. 15.3. Not surprisingly, the radiative collisions which had previously
been observed with 10 W of microwave power can be observed with
milliwatts of power, due to the high Qof the cavity. The most interesting
aspect of the data, though, is what happens instrong microwavefields.In Fig.
15.5 we showthe developmentof the resonances observed for the radiative
collision process of Eq. (15.5) as the 15.5 GHz microwave power incident
on the cavity is increased.8 With no microwave power only the four
collisional resonances corresponding to the different \m\values of the 17p
and 18p states are observed. When the microwave power is raised this
pattern of four resonances is repeated three times at lower static tuning
fields, corresponding to the stimulated emission of one, two, and three
microwave photons during the collision. A straightforward way of labeling
the resonances is simply to extend our previous notation to account for the
number of photons emitted. Explicitly, we label the collisional resonances as
(mhmu)m where mxand rau are the me valuesof the upper and lowerpstates
and m is the number of photons emitted. In Fig. 15.5 the (0,0)mresonances
are indicated by m with an arrow. It is hardly surprising that, as the
microwave power is raised, higher order multiphoton processes are
observed. On the other hand, it maybe surprising that for each m # 0 the cross
sections first increase then decrease with microwave power. For example,
the m = 1cross section is clearly zero in the lowest trace. Similarly, the m = 0
cross section vanishes in the trace one above the lowest but reappears in the
lowest trace. Such behavior, typical of the strongfieldregime,is notpredicted
byperturbation theory. Closeinspection ofFig. 15.5 revealsthat the positions
of the collisional resonances shift to lower staticfieldas the microwave
power is raised. Finally, in contrast to the usual observation of broadening
with increased power, the (0,0)m resonances, which are well isolated from
other resonances, develop from broad asymmetric resonances to narrow
symmetric ones as the microwave power is raised. That the observed
resonances for m # 0 become narrower and more symmetric as the power is
increased can be understood in the following way, using the onephoton
process asan example. The coupling matrix element of Eq. (15.2) hastwo

320 Rydberg atoms I _ 220 260 300 340 380 420 460 ELECTRIC FIELD
(V/cm) Fig. 15.5Observed Na 18pion signal after the population of the 18s
levelvsthe static field with a 15.4 GHz microwave field. Trace (a)
corresponds to no microwave power input to the cavity and showsthe set
offour zero-photon collisional resonances.Traces (b), (c),(d), and (e)
correspond, respectively, to 13.5, 50, 105, and 165 V/cm microwave field
amplitudes inside the cavity and show additional sets of four collisional
resonances corresponding to one, two, and three-photon radiatively assisted
collisions. The peaks labelled 0,1,2, and 3correspond to the
lowestfieldmember of the set of four resonances corresponding to zero-,
one-, two-, and three-photon assisted collisions, (0,0)�, (0,0)*, (0,0)2,
(0,0)3 (from ref. 8).

Theoretical description 321 factors, a dipole-dipole term, and a dipole-field

term. For the collision to be observable, the time integral of the matrix
element over the collision must exceed some minimum value ~1. Thus, if the
microwave field is smaller, the radiative collision must occur with a smaller
impact parameter to compensate. A small impact parameter hastwoeffects.
First, itimplies a short collision time and awide resonance. Second, as the
atomscomecloser together the molecular energy levels deviate from their r =
o� values, producing the asymmetry. As the microwave field is raised, the
impact parameter can become larger, and the resonances become narrower
and more symmetric. Theoretical description The experimental data shown in
Figs. 15.3 and 15.5 were obtained with a microwave frequency CDIITI > 1/r
where r is the time or duration of the collision and IIr is the linewidth. In this
case the resonances corresponding to the absorption or emission of different
numbers of photons are resolved. Here we describe radiative collisions
starting from the high frequency regime, (D/2JZ > IIT and progressing to the
low frequency regime, oolln < lit. The strbngfield',highfrequency regime To
describe the shifts and intensities of the m-photon assisted collisional
resonances with the microwavefieldPillet etal. developed apicture based on
dressed molecular states,3 and we follow that development here. As in the
previous chapter, we break the Hamiltonian into an unperturbed Hamiltonian
Ho, and a perturbation V. The difference from our previous treatment
ofresonant collisions is that now Ho describes the isolated, noninteracting,
atoms in both static and microwave fields. Each of the two atoms is
described by a dressed atomic state, and weconstruct the dressed molecular
state asadirect product of the two atomic states. The dipole-dipole
interaction Fis stillgivenby Eq. (14.12),and usingitwe can calculate the
transition probabilities and cross sections for the radiatively assisted
collisions. We begin by developing the wavefunction for one atom in
combined static and microwavefields,both of which are in the same
direction. If the nominal nt state has an energy Wn�(E) in astaticfieldE, we
can describe the energy in the vicinity of the staticfieldEs by Wn, (ES+AE) =
W'n� + kn<AE +�an<(AE)2 (15.6)

322 Rydberg atoms where ,wn and d2Wn dE2 If wenowimagine that AE
varies in time, but slowly,itsonly effect is to cause a time variation of the
energy of the ni state. We assume that the spatial wavefunction is unaffected
by A� and that no transitions occur. This approximation is the adiabatic
approximation of Autler and Townes.11 Now let us consider the time
variation of AEof particular interest to us, �"mw coscot. Ifthe assumptions
stated above are valid, we can use the energy of Eq. (15.6) as the
unperturbed Hamiltonian Ho in theSchroedinger equation. Explicitly, W(t)
yjn( (t) = (W'ne + kne Emv/ cos wt +^f (Emw cos cot)2)tpn( (t) = ^ % ^ ,
(15.7) which has thesolution W(0 = </we-|>'>d<'. (15.8) Carrying outthe
integrations and discarding the phase factors from t0 yields Vw(0 = VW e-K
+ T)' e - ' ^ i - , e-'*#**�*". (15.9) The exponentials with sinusoidal
arguments may be expanded as Bessel functions using12 00 ^ k(x) e ik �f.
(15.10) With the Bessel function expansion of Eq. (15.10), Eq. (15.9)
becomes (15.11) If an^mw2 � co, as is usually the casefor the experiments
described here,inthe last expansion ofEq. (15.11) only the k' = 0 term
contributes and the summation is equal toone. In this case the wavefunction is
written as CO (15.12) We note that for collisions involving states of good
parity, kn� = 0, andthe first summation of Eq. (15.11) collapses to 1. In this
case wemaynot disregard the second summation of Eq. (15.11). As discussed
in Chapter 10,wecan think of the effect of the microwavefieldon the nt state
asmodulating its energy. Just asmodulating the frequency ofa radio wave
produces sidebands, the microwave field canbe thought of as modulating the
energy of the nt state and breaking it into carrier and sideband states.

Theoretical description 323 ERG Y z LU +2 +1 ( 18p " ^ ^ " " ^ 18s 17p -***
" ) -1 -2 --� Fig. 15.6Energy levelsofthe Na 17p, 18s,and18p states
showingthefirstupper and lower sideband statesofthe pstates.Thenumbers
+2,. .., -2 refer to the net number of photons emitted in a radiative collision at
these tunings of the field. Note that there are several processes which lead to
thenet emission of, for example, zero photons (from ref. 3). In Eq. (15.12)
x/jn� is the spatial wavefunction of the nt state at the staticfield Es, W'n� is
the energy in the staticfieldEs, an�Emw2/4 is the acStark shift produced by
the microwave field, and Wne + an�Emw2/4 is the carrier energy. The
Bessel function Jk(kEmv,/co) gives the amplitude in the kth sideband state,
which is displaced in energy from the carrier state by kco, k microwave
photons. For the np states knp is large, but for the ns states kns isvery small,
and the amplitudes of the s sidebands are negligible. A graphic presentation
of the effect of the microwave fields is shown in Fig. 15.6, a plot of the
energies of the 18s, 17p and 18p states showing the carrier energies of these
states as well as the energies of the � 1 sideband states of the p states.3 As
mentioned above, and shown explicitly in Eq. (15.12), the amplitude of each
sideband depends on the strength of the microwave field, and while this point
is not made explicitly in Fig. 15.6, it is implicit in that no sidebands of the s
state are shown and second and higher sidebands of the p states are not
shown. Thus Fig. 15.6 corresponds to a microwave field too weak to have
observable amplitudes for the second sidebands of the p states. With no
microwaves all sidebands would have zero amplitude and the carriers
amplitude one. A further interesting point to observe inFig. 15.6 isthat there
are twoprocesses corresponding to the net emission of �1 photon, and three
corresponding to the net emission of zero photons. These processes
correspond to the transitions between the

324 Rydberg atoms energetically allowed combinations of sideband states,

raising the question of how to account for these different processes, all of
which occur at the same static field and are therefore experimentally
indistinguishable. Using the dressed atomic states we construct the dressed
molecular states as direct products. Explicitly, we construct the two states
ipA(t) an d VB(O given by3 VA(0 = Vi�s(O � VaisW (15.13a) and VB(0 =
W O � ^2(w-i)p(0 (15.13b) in which each of the individual wavefunctions
is given by Eq. (15.12). If we write out the expression for ipB explicitly
wefind A;A:' h ( ^ M 4'(fc("'fmj e^*'* (15.14) where WB = W'np + W[n-1)p+
(anp + a ^ ^ ^ w ^ , the sum of the carrier energies of the ftp and (n � l)p
states. With the Bessel function identity,13 00 jN(x�y)= Y -W44O0, (15-15)
the double summation of Eq. (15.14) may be condensed to a single sum,
yielding -**", (15.16) where kB = knp + k(n_^p. Similarly we can write the
wavefunction for ipA(t) a s 'A.| CO where WA = 2W'ns and kA = 2kns Both
V;A(0 an d V*B(0 a r e solutions to H0, and the total wavefunction is, in
general, a linear combination of the two. Explicitly, it is given by V(0 =
CA(0VA(0 + CBCOV'BCOJ (15.18) where CA(f) and CB(t) are time
dependent coefficients analogous to those in Eq. (14.15). If we insert the
wavefunction of Eq. (15.18) into the time dependent Schroedinger equation,
HV(t) = idV/dt, (15.19) and use the fact that ipA(f) and ipB(t) are solutions
of the unperturbed Schroedinger equation, we find the equation VCA (0VA(0
+ VCB(t)tpB(t) = i CA(t)yjA(t) + iCB(t)jpB(t). (15.20) Assuming that the
diagonal matrix elements of V vanish, multiplying Eq. (15.20) in turn by ip\
(t) and ipB(t) and carrying out the spatial integrations yields CB(0 (15.21a)

Theoretical description 325 and <VA(OMVB(O> C B � = iCA(O- (15.21b)

We note that Eqs. (15.21) are similar to Eqs. (14.16) but for the time
dependence of VA(0 an d V'BW- We recall that V is always time dependent
since it depends on the internuclear separation of the atoms. However, this
time dependence is slow compared to those of VA(0 a n d V>B(0 Our
interest isin computing the transition probability at resonance, a condition
which is met when WA= WB + mco where m isan integer. First, weexplicitly
write the matrix element (VBCOMV'ACO) using the wavefunctions of Eqs.
(15.16) and (15.17) and the Bessel function relation of Eq. (15.15). It is
given by \ co h ( < * A " * B > M e-^A - ^ + *�]'. (15.22) In this form it is
apparent that the matrix element (VBCOMV'ACO) *S a product of the spatial
matrix element, which varies slowly with time as the atoms move past each
other, and the Bessel function sum, which contains an enormous range of
frequencies. At resonance one term of the sum is constant in time and the
others oscillate at multiples of co. The oscillating terms are ineffective in
causing transitions in collisions of duration much longer than the microwave
period, since they average to zero. Thus at the resonance corresponding to
the radiative collision in which m photons are emitted, i.e. when WA= WB
4- mco, the constant and significant part of the matrix element of Eq. (15.22)
is given by (15.23) where K= kA � kB. The matrix element of Eq. (15.23) is
identical to the matrix element for a normal resonant collision multiplied by
aconstant factor, the Bessel function, and as aresult the transition probability
can be computed using the same approaches as in Chapter 14. All radiative
collision experiments with Rydberg atoms have been done with the collision
velocity perpendicular to the static and microwave fields. Thus the results
obtained are an average over spatial orientations, and it makes little sense to
use adetailed model of the interaction. Accordingly, we assume that y V5b
V5b where b is the impact parameter and % = jutju2.With this assumption,
the atoms interact for a time blv. Using matrix elements of the form of Eq.
(15.24) leads to

326 Rydberg atoms sinusoidally oscillating solutions of CA(f) and CB(t)

during the collision . If we choose as initialconditions CA(�oo) = 1and
CB(� o�) = 0,then after the collision, C\(i) =cos2 ^-/m ^ w j j , (15.25a)
and the transition probability at impact parameter b, Pm(b), is given by C|
(0Explicitly, Pm(b) = C2 (0 = sin2 [-f/m ( ^ M l . (15.25b) The transition
probability of Eq. (15.25b) is of the same form as the transition probability
of Eq. (20.25), with D = x\Jm(KEmw/a))\. Accordingly, ifwe define an
impact parameter bm(Emv/) such that X 11 2' Pm(b) oscillates between 0
and 1as b increases from zero to bm(Emw), at which point Pm(b) = 1 for the
last time. As b increases from bm(Emw) to infinity Pm(b) decreases
monotonically from 1 to 0. Toobtain the crosssection for the m-photon
assisted collision inthepresence of a microwave field Zsmw we integrate the
transition probability over impact parameter. Explicitly tfm(�mw) = JO
Carrying out the integral explicitly yields a (E ) =� b 2 (E ) = �% m mw ^
m mw ^ y KE mw 0) /� 0) db. (15.27) (15.28) A radiative collision with the
emission of no photons in the absence of a microwavefieldissimply aresonant
collision. SinceJo(0) = 1,we can express the radiative collision cross section
of Eq. (15.28) as (15.29) w i.e. asthe product of the resonant collision cross
section aR and the magnitudeof the Jm(KEmw/(o) Bessel function. Eq.
(15.29) is onlyvalid inthe strongfieldregime. It is clear that \iJm{KEmJa))
issmall,using Eq. (15.26) wecompute asmallvalue of bm(Emw). When
thisvalue approaches the size of the atom, Eq. (15.29) isno longer valid, for
the integralof Eq. (15.27) is dominated by contributions for b ~ bm(Emw),
where neither the assumption of the energies being independent of rnor the
dipole approximation

Theoretical description 327 are valid. Under such circumstances a collision

with a small impact parameter does not lead to as large a transition
probability as indicated by Eq. (15.25b). Connection to the weakfieldregime
In the experiments with Rydberg atoms it is very difficult to observe
radiatively assisted collisions with cross sections more than a factor of 10
smaller than the resonant collision cross sections, so the deviations from Eq.
(15.29) are not apparent. However, in other contexts, such as laser assisted
collisions, this limitation does not apply, and it is interesting to consider how
the above description passes over into the weak field regime, in
whichJm(KEmJw) is small. If we restrict the integration in Eq. (15.27) to the
large rregion of space, in which the approximations we have used are valid,
we can rewrite Eq. (15.27) as Om = KEmw ' (15.30) in which an obvious
lower limit for B is the geometrical radius of the atom. Acccordingly, B =
n2a0 for a Rydberg atom of principal quantum number n. An important fact to
note about Eq. (15.30) is that om(Emw) is proportional to J2m(KEmJa)).
Using the fact that for small arguments Jm(x) = (x/2)m,12 at low microwave
fields <*/�w, (15-31) i.e. the cross section for an ra-photon assisted
collision is proportional to the microwave intensity to the mth power, as
expected on the basis of time dependent perturbation theory. The
lowfrequency regime In the previous discussion of radiatively assisted
collisions we have assumed that there are many cycles of the radiation field
during the time of one collision, i.e. col In � IIT. It is interesting to consider
the opposite extreme, OOI2TC � IIr. We consider collisions in the total field
E = Es + �rf cos (cut + 0), (15.32)

328 Rydberg atoms and we assume that the collisional resonances are
observed by scanning the static field. If, inthe absence of the rffield,the
resonance isfound at Es = ER, itisfound at Es = ER - �rf cos (cot + 0)
(15.33) with the added rffieldof Eq. (15.32). If wecan control the phase of
the rffieldat whichthe collision occursthe resonance is simply shifted as
shown byEq. (15.33). If we cannot control the phase, either because the
exciting laser is not synchronized with the rf field, or because collisions can
occur randomly over several rf cycles, then the probability of observing the
resonant collision signal at any given staticfieldis proportional to the
likelihood of the totalfield'sbeing equal to EK at that staticfield.If the
resonant collision crosssectionsis given as a function of field byoR(E), then
in thefieldEs + Erf cos cotthe crosssection observed at Es is given by the
convolution of oR(E) with the function f(E)JjzETf[l-(E-Es/ETf) 2 ]1/2
\E'Es\^En ( i 5 3 4 ) 10 \E-ES\>ET{ In other words o(Es) isgiven by the time
weighted average of the cross section in the static and
oscillatingfields.Sincethe oscillatingfieldspends most of its time at its turning
points, the observed cross section can be expected to exhibit the behavior
shown in Fig. 15.7.3 The intermediatefrequency regime We have considered
the high frequency regime, which is independent of the rf phase, and the
lowfrequency regimewhichis critically dependent onthe rf phase. We now
consider the intermediate frequency regime, CO/2JZ ~ 1/r, approaching it
from the low frequency side. To date the only way of calculating the
lineshape of the collisional resonances has been
explicitnumericalintegration. Aswe shall see, when comparisons with
experimental results are made, the calculations predict multiple peaks in the
cross section. The separation between peaks does not, however, correspond
to the rf frequency. We can use a simple picture to understand the origin of
the peaks and their locations, although the picture does not give
lineshapes.14 For concretenesswe assumethat thecollisionoccursoverthe
timeinterval � 772 <t<T/2 in the presence of arffieldof welldefined phase
sothat the totalfieldis given by Eq. (15.32). We consider the transitions
between the two molecular states A and B which are degenerate (at r = o�)
at the static field ER. We may assume with no loss of generality that, in the
vicinity of ER, B has no Stark shift and A has a linear Stark shift. Ignoring the
dipole-dipole coupling, the energy,

Theoretical description 329 (a) (c) W(FS) Fig. 15.7 Collisions in the
presence of a lowfrequency microwave field: (a) the resonant collision
profile without themicrowaves; (b) oneperiod of microwavefieldshowing
both the amplitude and the time weighting ofthe extremefieldvalues; (c)
typical time weighted radiative collision lineshape when the microwave
period becomes long compared tothe duration of one collision (from ref. 3).
WB, of molecular state Bisaconstant, and the energy of molecular state A
isgiven by WA=WB + kA(E - ER), (15.35) where Eis given byEq. (15.32).
Inspecting Eq. (15.35) wesee that WA follows the rf field. Imagine that we
choose the rf phase <p = 0so that WA reaches a maximum in the center of the
allowed collision time. In Figs. 15.8(a) and (b), we show the energies WA
and WB as functions of time for two choices of the phase of a 0.75 MHz rf
field, 0 = 0 and jz/2, respectively. The allowed collision time T =0.7 jus.
The dipole-dipole interaction lifts the degeneracy at resonance, leading to the
avoided level crossings shown by the insets of Fig. 15.8. In Fig. 15.8(a) the
energies are shown for twovalues of the staticfieldEs. For ES<ER- Erf, the
two levelsnever come into resonance. When Es = ER � ETfthe twolevels
are resonant at t = 0, as shown bythe broken line, and a broad resonance can
be expected. For larger Es the two levels come into resonance twice,
asshown bythe solid line, and the transition amplitude isthe coherent sum of
the amplitudes acquired at the two times the levels become resonant, at t =
�f. As shown by Fig. 15.8(a), if the system is initially in state A it traverses
two avoided level crossings to reach the final state B, and there are two
possible paths to go from A to B. Whether the transition amplitudes for these
two paths add

330 Rydberg atoms (b) wB CD \ \ wA -0.25 -t1 0 t1 0.25 -0.25 0 0.25 time
(ps) Fig. 15.8The initial and final state energies WA and WB in the total
electric field E = Es + Eri cos (cot + 0)where the frequency CO/2JI = 0.75
MHz field and the allowed collision time T=0.7 JUS. WB isfieldindependent
and WA varieslinearly with the field, (a) The phase 0 = 0. No resonance
occurs for Es < ER - Eti. As the static field is increased, when Es =ER �
Er{ ( ) the twolevels come into resonance at t = 0. When Es >ER - Exi (�)
the two levels come into resonance twice, and there are two resonant
interaction periods att = �t'. The levels donot actually cross buthave
avoided crossings asshownby the insets. There aretwopaths bywhich atoms
initially in state A can reach state B, and they interfere constructively when
<�B, thephase difference accumulated between �t' and t', is 2JTN. In
(b)where 0 = jt/2andEs ~ ER, only one interaction period occurs for all
values of Es between ER � Ed andER + Exi (after ref. 14). contructively
ordestructively depends on the phase difference 4> between the two paths.
<E> issimply the area between the two curves inFig. 15.8(a), i.e., <&=("'
(wA-WB)dt. (15.36) J If 0 = 2JZN, where Nis an integer the interference is
constructive, whereas if O = 2JZ(N + 1/2) it is destructive. As Es is raised O
passes through even and odd multiples of 7t leading to constructive and
destructive interference in the cross section. Inspecting Fig. 15.8(a)
wecanseethat thefringes should become closer together asEs is increased.
This description ofthe origin ofthe oscillations in the cross section makes it
apparent that they are a form of Stiickelberg oscillations observed in
differential scattering.15 The two resonant interactions at t = �t' can also be
thought of as being analogous to the two separated oscillating field regions in
a Ramsey magnetic resonance experiement, inwhich case the oscillations
inthe cross section can be thought of as Ramsey fringes.16 For 0 = 0,
asshown inFig. 15.8(a) the oscillations in the cross sections should extend up
to the staticfield ES =EK- Erf cos (a;772). (15.37)
Theoretical description 331 With the rf phase 0 = 71, the collisional
resonance is reversed with respect to EK. With (j) =jt/2, asshown in Fig.
15.8(b) for Es ~ EK, the result is qualitatively different. The two levels
come into resonance at most once for any value of Es, and as aresult there
isone broad collisional resonance extending from ES=EK ET{ sin(o>r/2) to
Es = EK + Erism(coT/2). The above description is obviously an extension of
the low frequency description, but it can also be connected to the high
frequency limit. Consider the casein which cu � 1/T, i.e., there are many
cycles of the field, and \E S � EK\ � � rf, the statictuningfieldis near
resonance compared to the rffieldamplitude. For a large transition
probability the transition amplitudes of successive rf cycles must add
constructively. This only happens when the phase difference <i> over an rf
cycle satisfies <E> = 2JTN with TV an integer. Using the energies of Eq.
(15.35), CJT/OJ <S> = kA(Es + Erf cos cot - EK) dt 0) Equivalently, kA(Es
- ER) = Nco. (15.39) In other words the resonances are detuned from ER by
the static field shift equivalent to Nphotons, the same result asobtained in the
dressed state picture. Eq. (15.39)tellsus where the collisional resonances
occur,but itdoesnot tell us how strong they are. The strength is determined
bythe phase contributions of the two parts of the rf cycle, when WA < WB
and WA > WB. When the phases accumulated in these two half cycles are
both integral multiples of 2n the collisional resonances are strong. This
requirement leads to an oscillation in the intensity of the N photon assisted
collisional resonance proportional to cos(kAETf/a) + y), where y is a small
constant, which has the same period as the Bessel function expression of Eq.
(15.29). Whilethe experiments withNa atomshavevalidated many
aspectsofthe theory of radiatively assisted collisions in the strong field, high
frequency regime, experiments with velocity selected K atoms have extended
the measurements both to higher order processes and through the regime in
which the collision time matches the frequency of the oscillating
field.141718 Specifically, studying the process K(w + 2)s + Knd-> Knp 4-K
(n - l)p + mco (15.40) with velocity selected atoms has several attractive
features. First, since the collisional resonances are so narrow, it is
straightforward to satisfy the condition O>/2JZ � lit yet still observe
collisions radiatively assisted by many photons without exceedingthe
availabletuningrange. Second, since thecollisiontimesare long, it ispossible
to observe the predicted phase dependence of the collisions in the transition
region r ~ 2JZ/(D.
332 Rydberg atoms The techniques employed are the same as those used to
study the K resonant collisions described in the previous chapter. The
apparatus shown in Fig. 14.13 is used, the only difference from the previous
experiment being the application of a rf voltage to the top plate of the
interaction region. The rf field, at frequencies from 0.5 MHz to 4.0 MHz is
either left free running or phase locked to the laser excitation. Most of the
data are acquired in the same way asdescribed above. Katoms are excited to
the 29s and 27d states by the laser excitation, 4s -> 4p-> 29s,27d. The atoms
are allowed to collide for 1jus, after which arapidly rising detuning pulseis
applied, followed by the more slowly rising field ionization pulse. Atoms
which have made the transition to the 29p state are selectively ionized by the
field ionization pulse and detected. This signal is monitored as the small
static tuning field is scanned. The amplitude and phase of the rf field are
changed as parameters. Using such low collision frequencies it is possible to
study radiative collision processes of high order, apoint made explicitly
byFig. 15.9.In Fig. 15.9 we show the development of the radiative collision
signal for collisions in the presence of4 MHz rffieldsof amplitudes from 0to
0.76 V/cm.17 Unlike the Na case described above, there is only a single
collisional resonance when there is no rf field, as shown by Fig. 15.9(a). As
the rf field amplitude is raised, progressively higher order resonant collision
processes are observed, out to the seven photon assisted collisions. It is
interesting to note that both m > 0 and m < 0radiatively assisted collisions
are observed, approximately symmetrically about the position of the
collisional resonance at 6.44 V/cm.A slight asymmetry is expected at the
extreme static tuningfieldsdue to the second order Stark shift. Close
inspection of Fig. 15.9 suggeststhat the strength of aparticular m-photon
assisted cross section oscillates with the rf field as implied by the Bessel
function dependence of Eq. (15.29). To demonstrate this point explicitly, the
static field wasfixedat thefieldscorresponding to several mphoton assisted
resonances and the amplitude of the rf field swept. Following this procedure
for m = 0, +1, -2, and +3 leads to the results shown in Fig. 15.10,which
shows quite clearly several zeroes in each of the om cross sections.
Furthermore, the experimental results agree almost perfectly with the
expression of Eq. (15.28) and the dependence obtained by simply
numerically integrating Eq. (14.16) using time dependent energies WA and
WB. Close examination of Fig. 15.10 reveals that the observed cross
sections at the highest rf field amplitudes are smaller than the calculated
cross sections. This discrepancy is attributed to the fact that the static fields
at which the collisional resonances occur shift slightly due to the second
order ac Stark shift as the rffieldis increased. Afinal aspect of these radiation
collisions, shown experimentally by Thomson et al.,17 is that the cross
section integrated over all the collisional resonances
increaseswiththerffield.Since therathcollisional resonancehas acrosssection
of o = oK\Jm(KEmJco)\ and ^mJ2m(x) = 1, in general, S om > aR.17

Theoretical description 333 +6 +4 +2 0 -2 -4 Electric Field (V/cm) Fig. 15.9

The n = 29 resonance of K in the presence of a 4 MHz rf field. The rf field
amplitudes for thefivesets of data are: (a)0 V/cm, (b) 0.19 V/cm, (c) 0.38
V/cm, (d) 0.57 V/cm, and (e) 0.76 V/cm. The top axis labels the sideband
resonances according to the number of rf photons emitted by the atomic
system (from ref. 17). As we pointed out above, an attractive feature of the
velocity selected K collisions is that we can examine the region CO/2TZ <
1/r. We consider first the case in which there isno control over the phase of
thefieldat which the collisions occur. This case is exemplified by Figs.
15.11(b)-(d), which show the effect of adding progressively larger amplitude
1 MHz rf fields of uncontrolled phase. Since the laser fires at an uncontrolled
phase of the rf field, the observed crosssection can be calculated using Eq.
(15.34). As shown by Fig. 15.11 the cross section is broadened in
approximately the manner shown in Fig. 15.7. As shown by a

334 Rydberg atoms , 0.2 0.4 0.6 0.8 1.0 1.2 1.4 RF Field Amplitude (V/cm)
Fig. 15.10 Crosssectionsas afunction ofrffieldstrength for thefirstfour
ordersof sideband resonancesof the K29s+ K27d radiativecollisionsina
4MHz rffield,(a)The zero-photon resonant collision cross section, (b) the +1
sideband resonance, (c) the �2 sideband resonance, and (d) the +3 sideband
resonance.The solidlineshows theexperimental data, thebold lineindicates the
prediction the Floquet theory, and the dashed lineis the result of numerical
integration of the transition probability (from ref. 17). comparison of Figs.
15.11(d) and 15.11(e),obtained with 1 and0.5 MHz0.2V/cm rf
fields,respectively, the broadening of the collisional resonance is
independent of thefrequency ofthe appliedfield,depending
onlyonitsamplitude, aspredicted by Eq. (15.34).18 When the phase of
alowfrequency rffieldis controlled the observed collisional resonances
change dramatically. With thefieldofEq. (15.32), when the atoms are allowed
to collide during the interval -0.5//s<O<0.5//s with 0 = 0 or jr, we observe
the collisional resonances shown in Figs. 15.12(b) and (c) respectively, in
agreement with Eq. (15.33). With no rf field the resonance occurs at Es =
6.44

Theoretical description 335 6.2 6.4 6.6 Electric Field (V/cm) Fig. 15.11 The
K 29s + 27d resonance inthe presence of alowfrequency rffield.In zero rf
field (a), the FWHM is 1.6 MHz. In (b)-(d), a 1.0 MHz field of strength 0.05
V/cm, 0.1V/cm, and 0.2 V/cm respectively is present. The solid line in (b) is
a numerical integration of thetransition probability, andtheboldline
istheconvolution of aLorentzian lineshape with asinusoidal shift from
resonance. In (e), the rf frequency is 0.5 MHz andits strength is 0.2 V/cm. For
these low frequencies, the features are no long frequency dependent but
rather arefieldstrength dependent (from ref. 18). V/cm, as shown by Fig.
15.12(a). When 0=0 the collision occurs at the maximum of the rf field, and
the resonance occurs at a lower static field, as shown in Fig. 15.12(b).
Similarly when the collision occurs at the minimum of the rf field the
resonance occurs at a higher static field, as shown in Fig. 15.12(c). Finally
we consider the case in which (D/2JZ ~ 1/r. In Fig. 15.13 we show the
resonances obtained with an allowed collision time T = 0.8 jus (the
collisions are

336 Rydberg atoms Electric Field (V/cm) Fig. 15.12(a)The K 29s + K

27dresonance with a FWHMof 2.0 MHz. In (b) and (c), a 0.5 MHz, 0.1 V/cm
rffieldispresent with aphase of 0and nrespectively. A phase of 0 means the
cosine wave is at maximum during the temporal center of the collision (from
ref. 18). allowed tooccur for �0A JUS <t<0Ajus andtwo choices ofthephase
of anrf field, Er{cos(cot + 0), of amplitude Er{ = 0.21V/cm and frequency
0.75 MHz.14 InFig. 15.13(a) we show the resonance with no rffieldat 6.44
V/cm. In Fig. 15.13(b),in which 0 = 0, the experimental curve exhibits a
broad peak at Es = 6.27 V/cm, a subsidiary peak at 6.35 V/cm, and a
possible third subsidiary peak at 6.40 V/cm. The dotted linesarenumerical
calculcations andthelocationsofthepeaksfor N = 0-2 from Eq. (15.36)
areshown aswell. As can be seen, the locations of the experimental peaks
agree well with both the numerical calculation and the interference picture of
Fig. 15.8(a) which leads to Eq. (15.36). When 0 = nil, as shown inFig.
15.13(c), a single broad resonance is observed, with no apparent
Theoretical description 337 Detuning (MHz) -10 -5 0 5 6.20 6.30 6.40 6.50
6.60 Static Electric Field (V / cm) Fig. 15.13The experimental (�) and
calculated ( ) lineshapes of the n=29 radiatively assisted resonance of K as
functions of the staticfieldZ%. (a) The unperturbed resonance with no rf field
present, FWHM = 1.4 MHz. (b), (c) The resonance in the presence of a 0.75
MHz, 0.21 V/cm rf electricfield and a phase of (b) 0 = 0 and (c) 0 = jr/2.
Indicated below the spectra are the calculated positions of the peaks using
the <l> = 2JTN condition (fromref. 14).

338 Rydberg atoms Detuning (MHz) -10 -5 o 5 10 03 c CO CL G) CM (a)

<H> N=0 1 2 3 4 5 6 1,3 2,2 3,1 -,0 I I I I I I 5.9 6.0 6.1 6.2 6.3 6.4 6.5 6.6
6.7 6.B 6.9 7.0 Static Electric Field (V / cm) Fig. 15.14 The experimental
(�) and calculated ( )lineshapes asfunctions of the static field Es in a0.31
V/cm, 1.48 MHz, rffieldof phases (a) 0 = 0 and (b) 0 = n/2. The allowed
collision time is 0.8JUS. The calculated fringe positions using <l> = 2JIN
are indicated below the spectra. In (b) the fringe positions due to interference
between the first and second (solid vertical marker) and the second and third
(dashed marker) interactions are shown (fromref. 14).

References 339 subsidiary peaks. Asshown, the numerical calculations

agreewiththe experimental result. Since the levels are only degenerate once,
there is no possibility of interference, as shown by Fig. 15.8(b). As the rf
frequency is raised, interference becomes possible for all phases, a point
made by Figs. 15.14(a) and (b), showing the results for a 1.48 MHzfieldof
amplitude 0.31 V/cm and phases 0 = 0and 7t/2, respectively. The result for
<p = 7t is the mirror image of the 0 = 0 result. Asshown,theexperimental data
match the numerical calculation and the prediction of Eq. (15.36). In addition
to the fact that the 0 = jt/2 trace showsinterference peaks, we note that they
occur at approximately the same staticfieldas the 0 = 0 peaks, although they
are narrower. As the rf frequency is raised, the phase dependence slowly
disappears, until at 4 MHz it is undetectable, and we are in the high
frequency regime described by the dressed state picture. References 1. L. I.
Gudzenko and S. S. Yakovlenko, Zh. Eksp. Teor. Fiz. 62,1686 (1972) (Sov.
Phys. JETP 35,877(1972)). 2. A. Gallagher and T. Holstein, Phys.Rev. A
16,2413 (1977). 3. P. Pillet, R. Kachru, N. H. Tran, W. W. Smith, and T. F.
Gallagher, Phys.Rev.A 36, 1132 (1987). 4. R. W. Falcone, W. R. Green, J. C.
White, J. F. Young, and S. E. Harris, Phys.Rev.A 15, 1333 (1977). 5. P.
Cahuzac and P. E. Toschek, Phys.Rev. Lett. 40,1087. 6. A. V. Hellfeld, J.
Caddick, and J. Weiner, Phys.Rev. Lett. 40,1369 (1978). 7. W. R. Green, M.
D. Wright, J. Lukasik, J. F. Young, and S. E. Harris, Opt.Lett. 4,265 (1979).
8. P. Pillet, R. Kachru, N. H. Tran, W. W. Smith, and T. F. Gallagher,
Phys.Rev. Lett. 50, 1763 (1983). 9. R. Kachru, T. F. Gallagher, F. Gounand,
P. Pillet, and N. H. Tran, Phys.Rev. A 28, 2676 (1983). 10. R. Kachru, N. H.
Tran, and T. F. Gallagher, Phys.Rev. Lett.49, 191 (1982). 11. S. H. Autler and
C. H. Townes, Phys.Rev. 100, 703 (1955). 12. M. Abramowitz and I. A.
Stegun, Handbook of Mathematical Functions National Bureauof Standards
Applied Mathematics Series No. 55(US GPO, Washington, DC, 1964). 13. F.
Bowman, Introduction toBessel Functions (Dover, New York, 1958). 14. MJ.
Renn and T.F. Gallagher, Phys.Rev. Lett.67, 2287 (1991). 15. D. Coffey Jr.,
D.C. Lorents, and F.T. Smith, Phys. Rev. 187, 201 (1969). 16. N. F. Ramsey,
Molecular Beams (Oxford University Press, London, 1956). 17. D. S.
Thomson, M. J. Renn, and T. F. Gallagher, Phys.Rev.A 45,358 (1992). 18. M.
M. Renn, D. S. Thomson, and T. F. Gallagher, Phys. Rev.A 49,409 (1944).

16 Spectroscopy of alkali Rydberg states In thefirsttwo chapters wehave seen

that the Na atom, for example, differs from the H atom because the valence
electron orbits about afinitesized Na+ core, not the point charge of the
proton. As a result of the finite size of the Na+ core the Rydberg electron can
both penetrate and polarize it. The most obvious manifestation of these two
phenomena occurs in the lowest �states, which are substantially depressed
in energy below the hydrogenic levels by core penetration. Core penetration
is a short range phenomenon which is well described by quantum defect
theory, as outlined in Chapter 2. In the higher �states the Rydberg electron
isclassically excluded from the core by the centrifugal potential �(� + l^r2
, and, as a result, core penetration does not occur in high �states, but core
polarization does. Since it isnot a short range effect, it cannot be described
intermsof a phase shift inthewave function dueto a small r deviation from the
coulomb potential. However, the polarization energies of each series of n(
states exhibit an n~3dependence, sothe seriescan be assigned a quantum
defect. Unlike the low � states, in which the valence electron penetrates the
core,measurements of the A�intervals of afewhigh�states enable us to
describe all the quantum defects of the high � states in terms of the
polarizability of the ion core. Irrespective ofwhether the quantum defect is
due to corepolarization or core penetration, itis important tohavegood
valuesfor these quantum defects as their values are important in determining
the Rydberg state wavefunctions and, as a result, the Rydberg atom
properties. Although the fine structure intervals are invisible on the relatively
coarse energy scale of Fig. 2.2, the spin orbit splittings of the alkali Rydberg
levels differ from the hydrogenic intervals in a systematic fashion. Not
surprisingly the fine structure intervals for high � states are hydrogenic, but
those of the low �levels differ radically from H. Weareinterested in
measuring alkaliRydberg energy levels toextract quantum defects and fine
structure intervals. There are two ways of going about these measurements.
The first is to make purely optical measurements from low lying states to
determine directly the quantum defects andfinestructure intervals of the
optically accessible Rydberg states. The second is to measure the intervals
between the Rydberg states by rf resonance or related techniques. For
example, measuring the intervals between several high �states
allowsustofitthe observed intervals to a core polarization model yielding the
quantum defects. 340

Optical measurements 341 The data can be represented either as quantum

defects for each fine structure series or as a quantum defect for the center of
gravity of the � level and a fine structure splitting. For the moment we shall
use the latter convention, although it isbynomeansuniversal.Explicitly, we
represent the energyofann(j state, where j = � + s and s is the electron spin,
as 1 n�j 2(n-de)2 ' (16.1) where S is an empirical constant. The center of
gravity of the n� state has the energy ^ = - ,,. * - (16.2) Optical
measurements Historically the energies of the low � Rydberg states were
measured by optical absorption oremission spectroscopy. Today optical
spectroscopy is still avaluable tool for acquiring spectra, the most useful
modern forms being Doppler free techniques. One example is the single
photon excitation of atoms in a collimated beam by asingle frequency laser
beam which intersects the atomicbeam at aright angle. While the Doppler
broadening can never be eliminated due to the divergence of the atomic
beam, it can be substantially suppressed by this approach. For example,
Fredriksson etal.* excited Cs atomsin acollimated beam from the ground
6s1/2state to the7p3/2state using a4555 A laser, and 11% and1% of the
atoms in the 7p3/2 state decayed to the 5d5/2 and 5d3/2 states respectively.
From the 5dy states the atoms were excited to the nf5/2 and nill2 states by a
single mode dyelaser beam crossing the atomicbeam at aright angle.The
production of ni atoms was monitored by observing the ni-
5Afluorescence.They observed 10 MHz wide lines, yielding excellent values
for thefinestructure intervals of the Cs ni states. The second Doppler free
opticaltechnique istwo photon spectroscopy of atoms moving randomly in a
cell. The basic notion of two photon spectroscopy is straightforward. A laser
beam is retroreflected upon itself, usually so as to produce coincident foci of
the beams propagating in both directions in the center of the cell. In its own
reference frame an atom moving to the right with velocityv sees photons
coming from the right as having frequency v(l + v/c) and photons comingfrom
the left ashavingfrequency v{\ � v/c), where vis thelaser frequency and c
isthe speed of light.The sum of the twofrequencies is2v, irrespective of the
velocity of the atom. Thus all the atoms,irrespective ofvelocity, are equally
likely to absorb one photon from each beam when 2v equals an allowed two
photon

342 Ar* LASER \ - * DYE LASER \ TO FREQUENCY AND

WAVELENGTH y MONITORS TO RFrnpnFC Rydberg atoms OPTICAL \
ISOLATOR n An ' 4 I CHOPPER REF LOCK-IN AMP OVEN PYREX TUB
DETECTOR � � - - �! j r \ < <> �>E . - � � I I 1 T MIRROR Fig. 16.1
Schematicdiagram ofthe experimental arrangement for two photon
spectroscopy of Rb Rydberg states (from ref. 5). interval. The narrow two
photon absorption is accompanied by a much weaker Doppler broadened
absorption due to two photons from either beam. This method has been used
quite successfully to measure the two photon absorption spectra of K2"4,
Rb5'6 and Cs7'8. The most often used detection approach is showninFig.
16.1. The atoms arecontained in aglasscell,andthetwo photon absorption
isdetected bycollisional ionization of the Rydberg atoms. The cell shown in
Fig. 16.1 is usually called a space charge limited diode.9 The basic
principle of the diode isthat electrons are thermionically emitted from
ahotwire, but the biasvoltage on the wire issolowthat the current is limited
bythe electron spacechargenear thewire. If an atomisexcited to aRydberg
statein afairly dense vapor, it is collisionally ionized, and the ion formed is
attracted to the wire emitting the electrons. Near the wire the ion partially
neutralizes the electron space charge allowing a larger electron current
toflow.Since the ion moves very slowlyrelative to an electron, many
electrons escape through the holeinthe space charge created byeach ion
reaching the wire. Gains of 106have been reported for such diodes, and, not
surprisingly, they have found wide application in optical spectroscopy of
Rydberg states. The optical data give the energies of the members ofthe
Rydberg series relative to some lower state, not relative to the ionization
limit, and it isconvenient to use theterm energy Tn�.oftheft�ystate. It is the
energy abovetheenergyofthe ground state. Knowing the energy of the initial
state of the optical transition, all the observed transition frequencies in an
optical spectrum are readily converted to term energies. To extract
thefinestructure intervals is straightforward. They are simply the differences
in the term energies of the same n and �but different /. To extract the
quantum defects requires that we also determine the ionization limit. We
define the ionization limit IP asthe binding energy of the ground state, or
equivalently, as the term energy of an n = o� state. A straightforward
graphical

Radiofrequency resonance 343 approach can be used to determine both the

ionization limit and the quantum defect of the observed Rydberg series. When
the fine structure energies are removed, following Eqs. (16.1) and (16.2), the
resulting observed term energies, Tn�, of the centers of gravity of the n(
levels are related to the quantum defect de and the ionization limit IP by Tn�
= IP + Wn� = IP- 1 , (16.3a) 2(n - d�)z or in conventional unitsby T _ ,D R
y (n (16.3b) where Ry is the Rydberg constant for the atom under study.
Assuming that (5�� n we expand the right hand side of Eq. (16.3b) in
powers of <5�/n,which, when the higher order terms are ignored, leads to If
we simply plot Tn� + Ry/n2 vs 1/n3 we obtain a straight line the y intercept
of which gives the ionization limit and the slope of which gives the quantum
defect. In Fig. 16.2 this procedure is applied to the Li 5p to 15p term
energies from Moore10 to determine the ionization limit and quantum defect
of the Li np series. The ionization limit from Fig. 16.2 is 43,486(1) cm"1,
and the np quantum defect is 0.046(1). Radiofrequency resonance The
techniques described above allow accurate determination of the energy
levels of optically accessible Rydberg series. To measure the intervals to
other Rydberg series the primary technique has been microwave resonance.
The basic notions are identical to those used in molecular beam resonance
experiments. In the alkali experiments a tunable laser is used to excite an
optically accessible Rydberg state, which isthen driven to another Rydberg
state by amicrowave field which isslowly swept through resonance with the
atomic transition. To detect that the transition has taken place requires a
method of selectively detecting thefinal state of the microwave transition.
Since initial and final states of the microwave transition are necessarily
energetically close to each other, discriminating between them is not always
easy. Why microwave resonance techniques are so attractive for the
spectroscopy of Rydberg states becomes clear when we estimate how much
microwave power is required to drive the transitions. To drive the electric
dipole nd��ni transition A

344 Rydberg atoms 43500 43400 0 0.002 0.004 0.006 0.008 1/n3 Fig.
16.2Use of the term energies of the Linp Rydberg series to determine
theionization limit andthe quantum defect. Tnp + Ry/n2 for the5pto 15pstates
isplotted vs 1/n3. The y intercept is the ionization limit andtheslope is
�2Ryd p. �> B in atime r requires that the Rabi frequency beequal to
theinverse of the time r, or that juE = JT/T, (16.5) where// isthe dipole
transition moment and E is the microwave electric field. In a Rydberg atom
transition in which An = 0 and A� = 1,ju~ n2, and the exposure time of the
atoms to the microwaves can be several microseconds (~10n au.). Thus for n
~ 20,an electric field of ~ 10~13issufficient to drive thetransition. This is a
field of 500//V/cm, corresponding to a power density of 10~n W/cm2. The
fact that the single photon transitions require so little power prompts us to
consider two photon transitions. Consider the Na 16d�> 16g transition via
the virtual intermediate 16f state which is detuned from the real intermediate
16f state as shown by theinset of Fig. 16.3. If the detuning between thereal
and virtual intermediate states isA and the matrix elements between the real
states are//!and ju2, the expression analogous to Eq. (16.5) for a two photon
transition is

Radiofrequency resonance 345 4157 A ' 3 d 5890 A Fig. 16.3 Relevant

energylevelsfor theobservation ofthetwophoton, d-g resonance inthe n =
16state of Na. The straight arrows are the two laser pumping steps; the wavy
arrows downindicate the mostprobablefluorescentdecayofthe 16g
state.The3d-3p fluorescence at 8197 A is observed. The inset shows the
location of the d, f, and gstates, as well as the virtual f' state through which
the two photon process proceeds (from ref. 12). = 1 (16.6) A r WhileEq.
(16.6) depends ontheprecise detuning,ithasnot been uncommon to have
detunings of 10~2n~3. If we assume that//! = ju2 = n2, A = 10~2n~3,and r =
1011, when we evaluate Eq. (16.6) for n = 20 we find E ~ 10"11, or 50
mV/cm, corresponding to 1 juW/cm2. A very simple method of observing the
microwave transitions is to detect the fluorescence from only the final state.
Gallagher et al.11'12 used this technique to detect one, two, and three photon
Na transitions from the optically accessible nd states to the 3 < � < 5 states.
The scheme used to detect the Na 16d-*16g transitions is shown in Fig. 16.3.
Two pulsed dye lasers at wavelengths of 5890A and ~ 4150 A are required
to excite the Na nd states, which decay primarily by emitting a 4150
Aphoton. All of the �> 3 states arelikelyto decayvia the 3d state which
yields an 8200 A photon when it decays to the 3p state. Monitoring the
fluorescence at 8200 A with a color filter and a photomultiplier allowed the
microwave transitions from the d states to �>3 states to be detected easily.
A typicalresonance is theNa 16d3/2-16g7/2twophoton resonance
showninFig.16.4. There are several considerations to bear in mind when
using fluorescence detection. First, the approach is mostuseful
whenthephotonstobedetected have a vastly different wavelength than the
exciting light and the most probable decay of the optically excited state,
which need not be the same. Second, the branching ratio for the detected
transition should be favorable. Third, the lifetimes of the initial and final
state of the microwave transitions must be taken into account. If the
microwaves are always on, at resonance, radiative decay occurs from the
coupled pair of states. If the initial state of the microwave transition has a
much

346 Rydberg atoms 11530 11510 frequency (MHz) Fig. 16.4 Two photon Na
l6d3/2-^6g7/2 resonance on a sweep of increasing microwave frequency
(from ref. 12). faster radiative decay rate, few ofthe atomswill decayfrom
thefinalstate, and the resonance signal will be small. If the laser excitation
ispulsed and only apulseof microwaves is used, adifference in the radiative
lifetimes is unimportant. Finally, in a case such as the one shown in Fig. 16.4,
there is the possibility of a cascade decay from the n& state, via lower np
and ni states, to the 3d state. The cascade decay leads to a background signal
tending to obscure partially the resonance signal. Away of doingmicrowave
spectroscopy peculiar tothe studyof Rydberg atoms is to use selective field
ionization to discriminate between the initial and final states of the
microwave transition. An example of the application of this technique is the
measurement of millimeter wave intervals between Na Rydberg states by
Fabre etal.13 using the arrangement shown in Fig. 16.5. Using two pulsed
tunable dye lasers, Na atoms in a beam are excited to an optically accessible
ns or nd state as they pass between two parallel plates. Subsequent to laser
excitation the atoms areexposed tomillimeter wave radiation from a
backward wave oscillator for 2-5 jus, after which a high voltage ramp is
applied to the lower plate to ionize selectively the initial and final states of
the microwave transition. For example, if state A is optically excited and the
microwaves induce the transition to the higher lyingstate B, atomsinB will
ionize earlier inthefieldramp, as shown inFig. 16.5 . The A-B resonance is
observedby monitoring the field ionization signal from state B at tB of Fig.
16.5 as the microwave frequency is swept. In the method shown in Fig. 16.5,
the excitation, microwave transition and selectiveionization
alloccurinessentially the sameplace, sincethe thermal atoms donot movemore
than afew millimeters in a few microseconds. It is alsopossible

Radiofrequency resonance 347 Electron Multiplier Electric Field ulse Fig.

16.5 (a) Schematic diagram of the experimental setup, (b) Energy diagram of
the Na atom showing the levels populated by the stepwise laser excitation
and the microwave transitions, (c) Sketch of the sequence of events
experienced by the Na atoms (from ref. 13). to separate the three parts of the
experiment spatially, and an example is the measurement of the two photon
Na 39s �> 40s transitions by Goy et al.u They used two pulsed lasers to
excite Na atoms from the 3sto the 3p to the 39sstate as they entered a Fabry-
Perot microwave cavity. In the cavity, which is fed by a phase locked
backward wave oscillator, the microwave field drives two photon transitions
to the 40s state. After the atoms pass out of the microwave cavity they pass
between two parallel plates. A voltage, ramped in time, is applied to
selectively field ionize the Rydberg atoms which have made the transition to
the 40s state. Separating the microwave interaction region from the field
ionization region and allowing the atoms to interact with the microwaves for
a long time enabled Goyetal. to obtain resonance linewidths of �10 kHz. In
the discussion of fine structure measurements we give another slightly
different example of the use of selectivefieldionization to detect transitions
betweenfinestructure levels. Due to itshigh efficiency field ionization is a
very attractive technique, which is limited mainlyby the separation
intheionizationfieldsofthe levels in question. This issue is discussed in
Chapter 7. A third resonance method which has been used to measure the
intervals of alkali and Ba Rydberg atoms is delayedfieldionization which
takes advantageof the increase of the lifetime with �.The method used
bySafinya etal.15inthe study of Cs ni�>nh and ni�>nitransitions istypical.
Atoms are excited to the ni states in a manner similar to the one used by
Fredriksson et al.,1 except that pulsed

348 Rydberg atoms rather than cw lasers are used. The atoms are then
exposed to a 1 jus pulse of microwaves. After ~15//s, about three lifetimes
of the initial ni state, but lessthan one lifetime of the nh state, alltheRydberg
atoms are ionized. Atoms in the optically excited ni initial state have largely
decayed, but those which have undergone the ni �> nh transition have not,
and as a result, when the microwaves are tuned to a resonance the ionization
signal increases. The bigger the difference in lifetimes the better this method
works. The major limitation is that black body radiation redistributes
population from the initial state to other final states, producing asubstantial
nonresonant background signal.15 Other methods which have been used to
detect resonant microwave transitions between Rydberg states are selective
photoionization,16 and the detection of final states reached by black body
radiation from only one of the two states involved in the microwave
transition.17 A surprisingly accurate way of measuring the A� intervals is
the resonant collision approach used by Stoneman etal.18 to measure the
separations between the K ns, np, and nd states in small static electric fields.
Using the readily calculable Stark shifts of all the levels it isstraightforward
to determine the zero field energy levels of the p states relative to the well
knowns and d energies, which have been measured by Doppler free two-
photon spectroscopy.2'3 Core polarization The data from the microwave
resonance experiments give the differences in energies of the levels studied.
For the high �states, in which the difference in energy is due only to core
polarization, we can calculate the quantum defects themselves using a core
polarization modelfirstproposed by Mayer and Mayer.19 The model assumes
that the Rydberg electron is slowly moving in a hydrogenic orbit around an
ion core, which is polarized by the electrostatic interaction with the electron.
Slowly moving in this context means that the excited states of the ion core are
far removed in energy from theionic ground state compared to the intervals
between the Rydberg states. The polarization energy WpOlis given by Wpoi
= ^(r-4)-^(r-6), (16.7) where ad and aq are the dipole and quadrupole
polarizabilities of the ionic core. We can express the energy of the n� state
as Wn(= "i?+Wpo1 ^-T~2--r (16-8) In n

Core polarization 349 In Eq. (16.8) we follow the convention of Eqs. (16.1)
and (16.2) of separating the quantum defect from the fine structure. Values of
(r~4) and (r~6) for hydrogenic wavefunctions have been calculated
analytically, and the expressions are given in Table 2.3. Examining the forms
of Table 2.3, it is apparent that both (r~4) and (r~6) exhibit n~3 scalings, due
to the normalization of the radial wavefunction. However, they exhibit
different � dependences; (r~~4) scales as �~5, and (r~~6) scales as �~8.
As a result of the very different �scalings the contributions of the dipole and
quadrupole polarizabihties to the quantum defect are easily separated by
measurements of the intervals between several � series. Furthermore, for
high �, (r~4) � (r~6), and as a result, for high �, DOl "} ^ ' V * / or, in
terms of the quantum defect, <5< = | p - (16-10) It is a straightforward matter
to apply the description of Eqs. (16.7) and (16.8) to the observed energy
level intervals. In doing so we shall follow the convention of Edlen20 and
express the energies in wavenumbers using the Rydberg constant, Ry,
corrected for the nuclear mass of the atom under study. From Eq. (16.8) we
can see that for the � and V states of the same n the energy difference is
AW=WpoW,-WpoW, (16.11) which is implicitly a function of n, �, and �'.
We may write AW as AW = ad (Pn� - Pnr) + aq (Pn�Qn� - PnrQn�.),
(16.12) where Pne = Ry(r-\t, (16.13) and (16.14) In Eq. (16.12) AW is
expressed in cm"1, Pn� in cm"Vfl02, and Qn� in a02. Following Safinya et
al.15 we can write Eq. (16.12) more compactly as AW= ad AP + aq APQ,
(16.15) where AP = P w � - i V , (16.16) and AP2 = Pn(Qne - PnvQnv (16-
17) In Eqs. (16.15)-(16.17) AP and APQ are also implicit functions of n and
�.

350 Rydberg atoms 16.5 15.5 0 0.005 0.010 APQ/AP Fig. 16.6 Plot of
AW/AP vs APQ/APused to extract Cs+ effective dipole and quadrupole
polarizabilities, ad' andaq'. a'd isthey intercept of the line through the data
points while a' is its slope (from ref.15). Dividing Eq. (16.15) by
APremoves the n 3 scaling aswell as the contribution of the dipole
polarizability tothe observed A( intervals. Eq. (16.15) then becomes AW
APQ - ^ = � d + a q � (16.18) Eq. (16.18) consists of the measured
frequency intervals, AW, variables AP and APQ which are readily calculated
from the hydrogenic expressions in Table 2.3, the Rydberg constant, and the
unknown core polarizabilities ad and aq. Dividing the observed frequencies
AW by APand plotting them vs APQ/AP gives a straight line such as the one
shown in Fig. 16.6 for the Cs ng-nh-ni intervals. The y intercept gives the
Cs+ dipole polarizability, ad = 15.544(30)�o3, and the slope gives the Cs+
quadrupole polarizability aq = 70.7(20)ao5. If the Cs ni-ng intervals are
added to Fig. 16.6 they are found to be 20% too large to fit on the line,
indicating that core penetration contributes �20% to the quantum defect of
the Cs ni states. The � > 2 states of Na and Li have also been studied by
microwave spectroscopy, and the Na d states clearly exhibit core
polarization while the Li d states do not. In Table 16.1 we list the
experimentally measured15'21'22 core polarizabilities of the alkali ions Li+,
Na+, and Cs+ as well as the theoretically calculated values.23"27 We have
followed the convention of Freeman and Kleppner28 of expressing the
experimentally determined polarizabilities using primes, i.e. as ad and aq.
The primes serve to remind us that we have assumed that the outer electron is
very slowly moving. As discussed by Freeman and Kleppner, the fact that the
outer electron ismoving necessarily leadsto adeviation of a'd and aq from ad
and aq. However, as shown by Table 16.1, the agreement between the
measured values a'd and a'q and the calculated values ad and aq is quite
good, and

Table 16 Li+ Na+ Cs+ a (see ref. *(see c (see rf (see "(see '(see *(see ''(see
ref. ref. ref. ref. ref. ref. ref. i Corepolarization .1. Dipoleand
quadrupolepolarizabilities. measured \CIQJ 0.1884 (20)a 0.9980 (33)d
15.544 (30/ 21) 23) 24) 22) 25) 15) 26) 27) 0.046 (7)fl 0.35 (8)^ 70.7 (29/
Calculated ad (��) 0.189* 0.9459" 19.03* (&) 0.1122c 1.53" 118.26^ 351
the core polarization approach is a good way to describe the energies of
high� states. Thispoint is made graphically byFig. 16.7,a plot of alkali
quantum defects vs. �5.29 The linearity of the plot for high �shows that
high �states have quantum defects which are largely due to the dipole core
polarizability and that the quantum defects are well represented by Eq.
(16.10). Since the quadrupole contribution to the core polarization is small,
the deviation of the measured quantum defects from an (~5 dependence is
generally indicative of the onset of core penetration. From Fig. 16.7 it is
apparent that the Li np, Na nd, and Cs ni states are the highest � states to
have appreciable core penetration. As we shall see, thepresence or absence
of corepenetration playsanimportant roleinthe fine structure of the alkali
atoms. The quantum defects of alkali atoms have been measured far more
accurately than can be shown on a graph such as Fig.
16.7.2"8'10~15'21'22'30~35Accurate measurements reveal an energy
dependence of the quantum defects, which is conveniently parameterized
with a modified Rydberg-Ritz expression.34 The quantum defects are slightly
n dependent, and they are often specified for the resolved/ states, not their
center of gravity, asin Eq. (16.2). Explicitly, the quantum defect of an ntj state
is given by n<1 u (n-d0)2 (n-d0)4 (n-dof (n- dof ^ " The parameters (50, d2 .
. . are given in Table 16.2. For high n states the first two terms of Eq. (16.19)
are often sufficient. The term energies of these levels, Tn�j are readily
computed using Tnej = IP- Ry fk , (16.20)

352 Rydberg atoms s p d f g h i 10� 10-3 10-4 1 i I i r A 8 10~2 N \ \ \ 101

102 103 104 Fig. 16.7 Plot of the quantum defects d of the alkali atoms vs
�5: Li (�), Na (A), K (�), Rb (O), and Cs (�). Note the Vs dependence
for high � where the dipole polarizability dominates (from ref. 29). where
ityaik *s the Rydberg constant for the alkali atom. Table 16.3 is a compilation
of the ionization limits from the centers of gravity of the ground hyperfine
levels and the valuesofRya\k for the most common isotopes of the alkali
atoms. Fine structure intervals Measuring the A�intervals by microwave
resonance techniques generally yields the fine structure intervals aswell.
However, A� = 0transitions between the fine structure levels can also be
examined by several other techniques. The first of these is rf resonance.
Since the transition involves no change in � it is not an electric dipole
transition but rather a magnetic dipole transition, and a straightforward
approach is magnetic resonance, which has been used by Farley and Gupta36
to measure the 6f and 7ffinestructure intervals in Rb. Their approach is

Fine structure intervals 353 Table 16.2.Alkali quantum defectparametersfor

Eq. (16.19).a Series 7T � �_ "Pl/2,3/2 �d3/2,5/2 n t 5/2,7/2 23 Na
�s1/2c npl/2c np3/2c nd3/2 5/2 m 5/2,7/2 39 \C we Jv ^^1/2 ^Pl/2 �P3/2
�d3/2 nd5/2 "f5/2,7/2 85nu MC rvU ^^1/2 ^Pl/2 Wp3/2 �d3/2,5/2
^f5/2,7/2 1 3 3 rc we e La ^^1/2 "Pl/2^ �p3/2e �d3/2e �d5/2e nf5/2e
�f7/2e 60 0.399468 0.47263 0.002129 0.0003055(40) 1.3479692(4)
0.855424(6) 0.854608 0.015543 0.001663(60) 2.180197(15) 1.713892(30)
1.710848(30) 0.276970(6) 0.277158(6) 0.010098 3.13109(2) 2.65456(15)
2.64145(20) 1.347157(80) 0.016312 4.049325(15) 3.591556(30)
3.559058(30) 2.475365(20) 2.466210(15) 0.033392(50) 0.033537(28) d2
0.030233 -0.02613 -0.01491 -0.00126(5) 0.06137(10) 0.1222(2) 0.1220(2)
-0.08535 -0.0098(3) 0.136(3) 0.2332(50) 0.2354(60) -1.0249(10)
-1.0256(20) -0.100224 0.204(8) 0.388(60) 0.33(18) -0.59553 -0.064007
0.246(5) 0.3714(40) 0.374(40) 0.5554(60) 0.0167(5) -0.191(30) -0.191(20)
<54 -0.0028 0.0221 0.1759 0.7958 0.0759 0.16137 0.11551 -0.709174
-0.59201 1.56334 -1.8 -7.904 -0.97495 -1.50517 -0.36005 Se 0.0115
-0.0683 -0.8507 -4.0513 0.117 0.5345 1.105 11.839 10.0053 -12.6851
116.437 14,6001 -2.4206 3.2390 -0.206 -0.234 -2.0356 -26.689 -19.0244
-405.907 -44,7265 19.736 a (see ref. 34 unless stated otherwise)) ^(seeref.
21) c (seerefs. 35) rf (see ref. 22) e (see ref. 33) shown schematically in Fig.
16.8. Rb atomscontained in a cellareplacedin astatic magnetic field and
excited to the 5p3/2 state by an unpolarized resonance lamp. Then they are
excited to the nd5/2 state by a circularly polarized cw laser beam. Some of
the atoms decay from the nd5/2 state to the 6f and 7f states, and there is
enough polarization left in the f states that the detected nf-4d fluorescence has
a circular polarization. There is, in addition to the static magnetic field, an
oscillating magneticfieldof frequency �400 MHzperpendicular tothe static
field. The magneticfieldisswept, and whenitbringsaAraj= �1 transition
between the fine structure levels into resonance with the rf magnetic field it
reduces the polarization of the emitted fluorescence, allowing the detection
of the transition.

354 Rydberg atoms Table 16.3. Alkali ionization potentials and Rydberg
constant" atom 7 Li 23 Na 39 K 85 Rb 133 Cs Ryalk (cm"1) 109728.64
109734.69 109735.774 109736.605 109736.86 IP (cm"1) 43487.15(3)
41449.44(3) 35009.814(1) 33690.798(2) 31406.471(1) a (see ref. 34) RF
TRANSITION 5 S . TO PMT Fig. 16.8 Schematic illustration of the magnetic
resonance technique used to measure the Rb nifine structure intervals. The
Rb atoms in the n2F states are populated by spontaneous decay of the n'D5/2
states, which are populated by stepwise excitation of the ground state atoms.
The rf transitions, induced among the magnetic sublevels of the n2]? states,
are detected as a change in the intensity of the polarized /^F �> 42D
fluorescence. The lower part of the figure shows a sketch of the experimental
arrangement (from ref. 36). Both A; = 0 and A; = 1 transitions canbe
detected, the Ay = 1 intervalsbeing more sensitive to the fine structure
interval. Whilethefinestructure transitions areinherently
magneticdipoletransitions,it is in fact easier to take advantage of the large
A� = 1 electric dipole matrix elements and drive the transitions bythe
electricresonance technique, commonly used to study transitions inpolar
molecules.37In thepresence of a small static field of ~1 V/cm in the z
direction the Na nd; fine structure states acquire a small amount of ni
character, and it is possible to drive electric dipole transitions between them
at a Rabifrequency of 1 MHzwithan additionalrffieldof ~ 1 V/cm.

Quantum beats and level crossing 355 A good example of the useof the
electricresonance technique is the measurement of theNandfinestructure
intervals and tensor polarizabilities.38These transitions were observed
usingselectivefieldionization, although they appear tobeunlikely prospects for
field ionization detection because of the small separations of the levels, ~20
MHz. The nd3/2stateswere selectively excited from the 3p1/2state in a
smallstaticelectricfieldandthe Aray = 0 transitions tothend5/2 statesinduced
by a rf field parallel to the static field. When a slowly rising field ionization
pulse is applied to atoms in the Na nd3/2and nd5/2states of the same ra;, the
nd5/2 state atoms pass adiabatically to a high fieldstate inwhich \m\ = \m-\ �
1/2, whilethe nd3/2state atoms evolve to a statewith \m\= \mj\ + 1/2 as shown
bythe correlation diagram ofFig. 7.10. The lower mstates are more easily
ionized, and the peak of the ionizingfieldpulse isset so asto ionize the nd5/2
state but not the nd3/2 state of thesameray.When the radiofrequency is
scanned through resonance withthe fine structure interval asharp increase in
thefieldionization current isobserved.Due to the fact that the ionization fields
depend more strongly on m than on �, itis easier to observe these 20MHz
transitions between the ndfine structure states than it is to observe the 20
GHzd�> f transitions. The technique has also been used to measure Nanp
andK ndfinestructure intervals.38'39 Quantum beats and level crossing When
the fine structure frequencies fall below �100MHz they can also be
measured by quantum beat spectroscopy. The basic principle of quantum beat
spectroscopy is straightforward. Using a polarized pulsed laser, a coherent
superposition of the twofinestructure statesis excited ina time short compared
to the inverse of the fine structure interval. After excitation, the
wavefunctions of the two fine structure levels evolve at different rates due to
their different energies. For example if the nd3/2 and nd5/2 m} =3/2 states
are coherently excited from the 3p3/2 state at time t =0, the nd wavefunction
at a later time t canbe written as40 \ndjl) ={fl||i> c'iw^n + b\H) ew^t/2}
e^"172)'. (16.21) Here Wn isthe average energy ofthe twofinestructure
levels. The real numerical coefficients a and b, where a2+b2=ly depend on
the polarizations used in the excitation scheme, butare constant in time. Thus
the relative amounts ofd5/2 and d3/2statesdonot changewithtimebut
simplydecaytogether attheradiative decay rate F. However, the relative
amounts of m character oscillate at thefinestructure frequency, and this
oscillation ismanifested in any property which depends upon m, such as the
fluorescence polarized in a particular direction, or the field ionization signal
due to aparticular value ofm. This fact becomes more apparent

356 Rydberg atoms when we write the \jm) levels of Eq. (16.21) in terms of
the uncoupled \�m)\sms) states. Explicitly, 111) = V�|22>|i - i) +V||21> |H)
(16.22a) and || !> = Vf|22> |i - |) - V?|21) |U )� (16.22b) Written in terms of
the uncoupled states, Eq. (16.21) becomes | ndjl) = {A\22) |i - i> + B|21) |H>
|H> } e~r'/2, (16.23) where |A|2 = � (a2 + 462 + 4tffe cos WFS 0 and |�|2
= i (4a2 + b2- Aab cos WFSt). (16.24) As shown by Eq. (16.24), the amount
of ra = 2 and m = 1 character oscillates at the fine structure frequency. The
first measurements of Na nd fine structure intervals using quantum beats were
the measurements of Haroche et al,41 in which they detected the polarized
time resolved nd-3p fluorescence subsequent to polarized laser excitation for
n=9 and 10. Specifically, they excited Na atoms in a glass cell with two
counterpropagating dye laser beams tuned to the 3s1/2�> 3p3/2 and
3p3/2�> ndj transitions. The two laser beams had orthogonal linear
polarization vectors ex and e2 as shown in Fig. 16.9. The ftdj-3p3/2
fluorescence with polarization vector ed emitted at aright angle to the
directions of propagation of the two laser beams was detected. With ex 1 ed
and e2 1 ed the phase of the beat signal is reversed, as shown in Figs.
16.9(a) and 16.9(b). Subtracting the trace of Fig. 16.9(b) from that of Fig.
16.9(a) leads to the beat signalof Fig. 16.9(c). Fig. 16.9showsthe beat signal
for the Na9d state. The4 nspulse duration of the exciting laser and the 2 ns
risetime of the phototube set an upper limit of 150 MHz on the frequency of
the beat signals they could detect. Fabre et al.42 used the same technique to
extend the measurements up to n=16. An alternative method for detecting the
beats is field ionization, which is particularly useful for higher nstates. If
thefieldionization pulse isapplied rapidly compared to the fine structure
interval, the passage from the tsjm-}finestructure states to the uncoupled
tmsms states is diabatic and the fine structure states are projected onto the
uncoupled �rasrasstates, asrequired for the detection of a beat signal. Such
experiments were first carried out by Leuchs and Walther43 who applied a
rapidly rising field ionization pulse which projected Na nd atoms from the
coupled �sjm^ states onto the uncoupled �msms states and only ionized the
atoms in m = 0 and 1 levels but not those in m = 2 levels. Their observed
beat signals are shown inFig. 16.10. Using this technique they measured the
Na nd fine structure intervals over the range 21 < n < 31.

Quantum beats and level crossing 357 e2 g B2 XL. 'B, (c) Fig. 16.9
Recording of fine structure beats in the Na 9d level (averaging of 1000 runs)
obtained bytime resolved fluorescence detection. Trace (a):signalobtained
with configuration of polarizers e2 � e2>ed||e2 asshown in the upper inset;
trace (b): signal obtained with e1� e2,ed||e2 as shownintheinset;trace
(c):result ofsubtractingtrace (b)from trace (a). ex, e2, and ed are the electric
polarization vectors of laser beams 1 and 2 and the detected fluorescence,
and B1 and B2 are the propagation vectors of the two beams (from ref. 41).
Jeys etal.40 extended the measurements of Leuchs and Walther to higher n by
a taking advantage ofthe fact that theentirefieldionizationpulse need
notberapid, only the initial rise to a field large enough to uncouple the spin
and orbital motions. In fact, the requirements are quite modest. To project
thefinestructure states onto the uncoupled states arather smallfield,~1 V/cm,
isrequired. Jeys et al. applied a rapidly rising "freezing pulse" of ~1 V/cm at
a variable time after laser excitation to project the zero field fine structure
states onto the uncoupled states. The freezing pulse was followed by a
slowly rising ionizing pulse which allowed them to separate easily the
ionization of the diabatically ionizing m � 2 levelsfrom the adiabatically
ionizing m = 0 and 1 levels.Taking the ratio of the m = 2signal to the total
signal allowed them to record clear beat signals for n up to 40. Level
crossing spectroscopy has been used by Fredriksson and Svanberg44 to
measure the fine structure intervals of several alkali atoms. Level crossing
spectroscopy, the Hanle effect, and quantum beat spectroscopy are intimately
related. In the above description of quantum beat spectroscopy we implicitly
assumed the beat frequency tobehigh compared to the radiative decayrate
F.We show schematically in Fig. 16.11(a) the fluorescent beat signals
obtained by

358 Rydberg atoms QUANTUM BEATS n2D3/2-n2Ds/2 delay time Fig.

16.10 Quantum beat signals of high lying2D states of Na obtained by time
resolved selective field ionization. The variation of the beat frequency with
principal quantum number is shown. Several quantum beat frequencies
appear due to a Zeeman splitting of the fine structure levels in the earth's
magneticfield(from ref. 43). detecting the linearly polarized fluorescence
with two orthogonal polarizations. When WFS < T, the two
polarizedfluorescencesignals might be as shown in Fig. 16.11(b). Examining
Figs. 16.11(a) and (b) it is apparent that in Fig. 16.11(a), while there is a
phase difference between the intensities of the two polarizations, the
temporally integrated intensities are independent of WFS for WFS � F. On
the other hand, in Fig. 16.11(b) it is evident that when WFS < T not only are
the

Quantum beats and level crossing 359 detailed time dependences different
but alsothetemporally integrated intensities. The integrated intensity for
vertical polarization increases as WFS decreases and the integrated intensity
for horizontal polarization decreases. Wehave discussed
thefluorescenceintensity plots as iftheywereproduced by a short pulse of
light. In fact they also represent the probability of observing fluorescence
subsequent to excitation by a single photon. In a level crossing experiment
there isusually aweak but continuous stream of excitingphotons, and we have
nowayof knowingwhen anygiven photon excited the atom, soallwe can hope
to do is to measure the time integrated fluorescence. As we have already
seen, the integrated polarized fluorescence changes when WFS < F, and if we
could simply sweep WFS from WFS < �F to WFS > Fwewould see a
pronounced change in the integrated polarized fluorescence. In a level
crossing experiment the field dependence of two magnetic levels which
crossin a magneticfieldenables us tovarytheir energy spacing. In Fig. 16.8 we
show the energy levels of the Rb ni states in a magnetic field.44 The nf5/2
and nf7/2levels of different m}cross, andwithlightpolarized perpendicular
tothe field, o polarization, levels of m^ differing by 2 may be coherently
excited. Away from the avoided crossings there is no difference in the time
integratedfluorescence polarized parallel or perpendicular to the field.
However, at the level crossing there is a readily detectable increase in the
ratio of the parallel to perpendicular polarized fluorescence. From the field
at which the level crossing signal is observed and the known orbital and
spingf factors it isstraightforward to extract the fine structure intervals.
Fredriksson and Svanberg used a Na resonance lamp and atunable cw
dyelasertoexciteNa atomsinathermalbeamtondstatesviathe 3p states. They
scanned the static magnetic field to observe the level crossing signals for the
n = 5 to n = 9 nd states, most of which havefinestructure intervals too large to
be measured by quantum beat spectroscopy, and their signals for the lowest
field level crossings for n = 5to n=9 are shown in Fig. 16.12. To put the
alkali fine structure intervals in perspective it is useful to compare them to
the hydrogenic intervals. For H the energy of Eq. (16.4) is valid if45 6� = 0
for all �, and 5 where a is the fine structure constant. Using Eqs. (16.4) and
(16.25) we can express the hydrogenic fine structure intervals, the difference
in energy between the/ = � + 1/2 state and the/ =�- 1/2 state as WFS =
a2/2�(�+ l)n3, (16.26) i.e. the interval scales as n~3. In the alkali atoms
the interval scales only

360 Rydberg atoms 3/r Fig. 16.11 Intensity ofx( )andy ( )

polarizedfluorescenceas a function oftime after excitation withx polarized
light for different ratios of thefinestructure interval WFS to the radiative
decayfluorescenceF: (a) WFS = 10F, (b) WFS = F. In (a) there is only a
phase difference in the intensity of the emittedfluorescence.In (b) there is a
clear difference in the temporally integratedfluorescenceas well.

Quantum beats and level crossing 361 approximately as n~3. A better

representation of the intervals using the effective quantum number is (n-d) (n-
d) (n - 6) InTable 16.4 we tabulate the reported values of theparametersA, B,
and Cfor H and the alkali atoms.46"49 From Table 16.4several phenomena
are evident. First, the Liintervals and the nonpenetrating f states of Na have
very nearly hydrogenic intervals. Second the Table 16.4. Fine structure
parameters. Hnpa Hnda Hnfa Linpb Lindc Nanpd Namf Nanf' Knp8 Kndh Rb
npl Rb n$ A 87.65 GHz 29.22 GHz 14.61 GHz 74.53 GHz 29.096 GHz
5.376(5) Thz -97.8 (10) GHz 14.2(1) GHz 20.0847 (12) THz -1.131 (18)
THz 85.865THz 10.800 (15) THz Kbnik -152 GHz Cs/ip' Cs�dm 213.925
(20) THz 60.2183 (60) THz Csnin -979.6 GHz a (see ref. 45) *(see ref. 10) c
(see refs. 21 ^(see ref. 35) e (see ref. 47) f (see ref. 48) 8 (see ref. 18) ^(see
ref. 3) 1 (see ref. 49) 7 (see ref. 5) *(see ref. 36) 1 (see ref. 33) m (see ref.
8) "(see ref. 1) and 46) B 520 (20) GHz -6.78 (75) THz -15.53 (66) THz
-84.87 (10) THz +1.82 THz -56 (4) THz -58 (8) THz 12.22 THz C 13.05
(40) THz 90.4 (30) THz 390 (10) THz -33.76 THz

362 Rydberg atoms State 92D 82D 72D 62D 52D S ^ J-\ \ \ r A / V ^ 1771
1G Fig. 16.12 Experimental level crossing curves for the Na 5d to 9d states.
In each case the recording ismade at the lowestfieldAra; = 2 level crossing.
Each curve has been sampled for 1-2 h (from ref. 44). low�states for all the
alkalis but Lihave much largerfinestructure intervals than does H. Finally, the
Na and Knd states and the Rb and Csni states are inverted. These are the
highest �states of these atoms to exhibit core penetration. Two general
methods have been used to address the fine structure inversion theoretically.
One is to reduce the relativistic operators to a nonrelativistic form and use
them to calculate perturbations to the nonrelativistic solution. This general
approach has been used byHolmgren etal.50 and Sternheimer etal.51 The
other approach, used by Luc-Koenig,52 is to solve the Dirac equation in the
relativistic potential of the ionic core. Both approaches give similar
numerical results, and it has recently been shown by Lindgren and
Martensson53 that the approaches areinsomecases equivalent uptotermsof
order a2. From thepoint of view of a non relativistic calculation, the
inversion is due to polarization of the corebythevalenceelectron. Aspointed
outby Sternheimer etal.,for this effect to be strong, it is imperative that the
core have electrons with nonzero angular momenta. As a result, while the Na
nd states are clearly inverted it is not surprising that the Li nd fine structure
intervals deviate from the hydrogenic values by less than 1%.

References 363 References 1. K. Fredriksson, H. Lundberg, and S. Svanberg,

Phys. Rev. A 21,241 (1980). 2. C. J. Lorenzen, K. Niemax, and L.R. Pendrill,
Opt. Comm.39, 370 (1980). 3. D.C. Thompson, M.S. O'Sullivan, B.P.
Stoicheff, and G-X Xu, Can. J. Phys.61,949 (1983). 4. CD. Harper and M.D.
Levenson, Phys. Lett. 36A, 361 (1976). 5. K.C. Harvey and B.P. Stoicheff,
Phys.Rev. Lett. 38,537 (1978). 6. B.P. Stoicheff and E. Weinberger, Can. J.
Phys.57,2143 (1979). 7. C. J. Lorenzen, K.H. Weber, and K. Niemax, Opt.
Comm.33,271 (1980). 8. M.S. O'Sullivan and B.P. Stoicheff, Can. J.
Phys.61,940 (1983). 9. D. Popescu, I. Popescu, and J. Richter, Z. Phys.226,
160 (1969). 10. C.E. Moore, Atomic Energy Levels,NBS Circular 467 (U.S.
Government Printing Office, Washington, 1949). 11. T.F. Gallagher, R.M.
Hill, and S.A. Edelstein, Phys.Rev. A 13, 1448 (1976). 12. T.F. Gallagher,
R.M. Hill, and S.A. Edelstein, Phys.Rev. A 14,744 (1976). 13. C. Fabre, S.
Haroche, and P. Goy, Phys.Rev. A 18, 229 (1978). 14. P. Goy, C. Fabre, M.
Gross, and S. Haroche, /. Phys. B 13,L83 (1980). 15. K. A. Safinya, T.F.
Gallagher, and W. Sandner, Phys.RevA 22,2672(1981). 16. W.E. Cooke and
T.F. Gallagher, Opt. Lett. 4,173 (1979). 17. T.F. Gallagher, inAtoms inIntense
Laser Fields, ed. M. Gavrila (Academic Press, Cambridge, 1992). 18. R.C.
Stoneman, M.D. Adams, and T.F. Gallagher, Phys. Rev. Lett. 58,1324 (1987).
19. J.E. Mayer and M.G. Mayer, Phys.Rev. 43,605 (1933). 20. B. Edlen in
Handbuch der Physik, ed. S. Flugge (Springer-Verlag, Berlin, 1964). 21.
W.E. Cooke, T.F. Gallagher, R.M. Hill, and S.A. Edelstein, Phys.Rev. A 16,
1141 (1977). 22. L. G. Gray, X. Sun, and K.B. MacAdam, Phys.Rev. A
38,4985 (1988). 23. H.D. Cohen, /. Chem. Phys. 43, 3558 (1966). 24. J.
Lahiri and A. Mukherji, Phys.Rev. 141, 428 (1966). 25. J. Lahiri and A.
Mukherji, Phys.Rev. 153,386 (1967). 26. J. Heinrichs, /. Chem. Phys.52,
6316(1970). 27. R. M. Sternheimer, Phys. Rev. A 1, 321 (1970). 28. R.R.
Freeman and D. Kleppner, Phys.Rev. A 14, 1614 (1976). 29. T.F. Gallagher,
in Progress in Atomic Spectroscopy, eds. H.J. Beyer and H. Kleinpoppen,
(Plenum, New York, 1987) 30. C. Corliss and J. Sugar, /. Phys. Chem. Ref.
Data 8,1109(1979). 31. CJ. Lorenzen and K. Niemax, J. Quant, Spectrosc.
Radiative Transfer 22, 247 (1979). 32. K.B.S. Eriksson and I. Wenaker,
Phys.Scr. 1, 21 (1970). 33. P. Goy, J.M. Raimond, G. Vitrant, and S.
Haroche, Phys.Rev.A 26, 2733 (1982). 34. C. J. Lorenzen and K. Niemax,
Phys. Scr. 27, 300 (1983). 35. C. Fabre, S. Haroche, and P. Goy, Phys.Rev. A
22, 778 (1980). 36. J. Farley and R. Gupta, Phys.Rev. A 15,1952 (1977). 37.
H.K. Hughes, Phys. Rev. 72, 614 (1947). 38. T.F. Gallagher, L.M. Humphrey,
R.M. Hill, W.E. Cooke, and S.A. Edelstein, Phys.Rev. A 15, 1937 (1977).
39. T.F. Gallagher and W.E. Cooke, Phys.Rev. A 18,2510 (1978). 40. T.H.
Jeys, K.A. Smith, F.B. Dunning, and R.F. Stebbings, Phys. Rev.A 23,3065
(1981). 41. S. Haroche, M. Gross, and M.P. Silverman, Phys. Rev. Lett. 33,
1063 (1974). 42. C. Fabre, M. Gross, and S. Haroche, Opt. Comm. 13,393
(1975). 43. G. Leuchs and H. Walther, Z. Physik A 293, 93 (1979). 44. K.
Fredricksson and S. Svanberg, /. Phys. B 9, 1237 (1976). 45. H.A. Bethe and
E.A. Salpeter, Quantum Mechanics of One and Two Electron Atoms,
(Academic Press, New York, 1957). 46. J. Wangler, L. Henke, W. Wittman,
H.J. Plohn, and H.J. Andra, Z. Phys. A. 299, 23 (1981). 47. C. Fabre and S.
Haroche, in Rydberg States ofAtoms and Molecules, eds. R.F. Stebbings and
F.B. Dunning (Cambridge University Press, Cambridge, 1983).

364 Rydberg atoms 48. N.H. Tran, H.B. van Linden van den Heuvell, R.
Kachru, and T.F. Gallagher, Phys.Rev. A 30,2097(1984). 49. S. Liberman
and J. Pinard, Phys.Rev. A 20, 507 (1979). 50. L. Holmgren, I. Lindgren, J.
Morrison, and A.M. Martensson, Z. PhysikA 276, 179 (1976). 51. R.M.
Sternheimer, J.E. Rodgers, T. Lee, and T.P. Das, Phys.Rev. A 14, 1595
(1976). 52. E. Luc-Koenig, Phys.Rev. A 13, 2114(1976). 53. I. Lindgren and
A.M. Martensson, Phys.Rev. A 26, 3249(1982).

17 Rf spectroscopy of alkaline earth atoms The spectra of bound alkaline

earth atoms contain two features which are absent in the spectra of the alkali
atoms and He. First, there are perturbations in the regularity of the bound
Rydberg series caused by the presence of doubly excited states converging to
higher ionization limits. Second, in the higher angular momentum states there
is clear evidence for the breakdown of the adiabatic core polarization model
which we used in Chapter 16 to describe the alkali atoms.1 Although it may
not be apparent, these two features are slightly different manifestations of the
same phenomena. The perturbations of the spectra have been studied byboth
rf and opticalspectroscopy, but corepolarization canonlybe studied by rf
spectroscopy. Theoretical description of series perturbation and
nonadiabatic effects in core polarization For simplicity we shall consider the
Ba atom, although precisely the same procedure could be carried out for any
alkaline earth atom. We shall assume that Ba consists of an inert spherical
closed shell Ba2+ core which has two electrons.3 The basic notion of the
treatment of Van Vleck and Whitelaw2 is readily understood bylooking at the
energy level diagram of Fig. 17.1. Each one electron 0.2 G 0.1 i B 1 Fig.
17.1 Lowlyingenergy levels of Ba+ ( ), Rydberg series of Ba( )convergingto
them, andthe continua (///) aboveeachof them (from ref. 3). 365

366 Rydberg atoms state of Ba+ supports an entire system of bound and
continuum states of the outer electron. The bound 6sn� Rydberg
statesarethelowest lying system of bound states. As a first approximation we
can construct wavefunctions for the two electron states ignoring the
interaction between the two electrons. However, when the interaction is
introduced as a perturbation, there is a second order correction to the
energies of the bound 6sn� states which in most cases depresses their
energies. This description reduces, in the adiabatic limit, to the core
polarization model ofMayer and Mayer.1 The Hamiltonian forthe outer two
Ba electrons is given by } (17.1) r 12J where �f(r) is the potential ofan
electron ata distance r from the Ba2+ core and r12 is the distance between
the two electrons. As r�> o�,f(r) �> 21r. Separating H into H=H0 + H1
(17.2) leads to [y y ^] (17.3a) and yr2 + Vrl2. (17.3b) Using Ho, we can
write the time independent Schroedinger equation as +M) + +^] *(rr) =
W^(rr) (17.4) where rt and r2 are the vectors from the Ba2+ core to electrons
1 and 2. Eq. (17.4) may be separated into two independent equations, (17.5a)
l> J and ~ ( T + ~) ^(r2) = W2ip2(r2), (17.5b) where ^(ri?r2) =
V;i(ri)V;i(r2) (17.6a) and W= Wx+ W2. (17.6b)

Theoretical description of series perturbation and nonadiabaticeffects 367

The solutions to Eq. (17.5a) are the Ba+ wavefunctions (t>n'�'m'( r i), a n d
the solutions to Eq. (17.5b) are H wavefunctions un�m(r2). Explicitly, Eq.
(17.6a) becomes �(r!,r2) = 0�'�'m'(riK�m(r2), (17.7) and Eq. (17.6b)
becomes W=Wn,e,-l/2n2, (17.8) where Wn>z> is the energy of the Ba+ n'�
f state relative tothe Ba2+ ionization limit. The wavefunction of Eq. (17.7) is
an excellent representation of the high � Rydberg states even though it isnot
antisymmetrized and takes no account of exchange effects. From Eq. (17.8) it
is apparent that all states of the same n, �,and m are degenerate. This
degeneracy is removed by including H1 as aperturbation. For high �states
inwhich r2 > r1? the potential f(r2) can be written as f(r2) = 2/r2. Replacing
f(r2) by 2/r2 in Eq. (17.3b) and expanding l/r12 in terms of rx and r2 yields
Hx = � + � + � + � Px(cos <912) +1 P 2 (cos 012)... =r-�P1(cos e12) +
%P2(cos e12)..., (17.9) where d12 is the angle between rx and r2. It is useful
to rewrite the Legendre polynomials in terms of spherical harmonics.
Explicitly,4 P*(cos d12) = | ^ - j ^ 1^(01,0,)Yfan^fc) (17-10) where 0l5 01?
and 02,02 are the angular coordinates of electrons 1 and 2 relative to an axis
through the Ba2+ core. Since the angular eigenfunctions of Eq. (17.7) are
spherical harmonics and the ground state of Ba+ is an sstate, the diagonal
matrix elements of H1 vanish for the Ba 6sn( states, and there is no first
order correction to the energy from Hx. The second order corrections to the
energies of the Ba 6sn� states can beexpressed to adequate accuracy using
the dipole and quadrupole terms of Eq. (17.9). Explicitly, we write the
second order energy W2 as w � w + w (\ 7 11 \ r r 2 f r u r r q ? V / where
>*1 n'�'n"�") n'e'n"�" -�P,(cos012) t 6sn�] n't'ri't" (17.12)

368 Rydberg atoms and -I /6sn� iP2(cos612) n'e'n'T) (rit'n'T -�P2(cos012)

J �'�"�" (17.13) In Eqs. (17.12) and (17.13) summations over n' and n"
implicitly include the continua. The second order energy shifts given by Eqs.
(17.12) and (17.13) lead to both the series perturbation and core
polarization. We consider first the perturbation of the 6src�Rydberg series
whichoccurs when a low lyingBa5dn'T state liesin the middle of the 6sn�
Rydberg series. The 5d7d state is a good example of such a perturber. Due to
its proximity to the 6sn� Rydberg states the 5dnfT state dominates the sum of
Eq. (17.13) and the energy shift of a 6sn( level is given by 2 AW = Wq =
Sdn'T (17.14) The 6sn� Rydberg states lying above the
perturbing5dn"�"state are shifted up in energy, while those lying below the
perturbing 5dn"�" state are shifted down in energy. The Rydberg states lying
closest to the perturber are shifted the most. When there is more than one
perturbing levelit is often easier to use the quantum defect theory than to
apply Eq. (17.14). While the perturbation of the 6sn� Rydberg series
typically comes from the interaction with a single5dn'T state,
corepolarization comesfrom the interaction 0.2 rBa+6s Ba 6s18h I I I I I 0 1
2 3 4 5 6 <nCH/r2l18h>2(10-6a "4 Fig. 17.2 The squared
matrixelements(ngll/^llSh)2 ( )and <ni|l/^|l8h>2 ( ) plotted vs energy to show
the energy distribution of the squared matrix elements (from ref. 3).

Theoretical description of seriesperturbation and nonadiabatic effects 369

with entire Rydberg series converging to excited states of Ba+, i.e. many
terms of the sums of Eqs. (17.12) and (17.13) are important. The sums do not
extend over allpossiblevalues of n', �', n\ and�".With the help of Eq.
(17.10)we can see that in the expression for Wd, Eq. (17.12), �'=1 �"= �
� 1, and that in the expression for Wq9 Eq. (17.13), V = 2 and f = �, � �
2. In Ba, the 6s-6p dipole matrix element and 6s-5d quadrupole matrix
elements are by far the largest matrix elementsinthe summations ofEqs.
(17.12) and (17.13), and theseterms also have the smallest energy
denominators. Thus we can, to an excellent approximation, restrict the sum
over n' to one term in each case. The moststraightforward wayof evaluating
the angular matrixelementsofEqs. (17.12) and (17.13) is to use the method of
Edmonds.5 The matrix elements are evaluated as the scalar products of
tensor operators operating on the wavefunctions of electrons 1 and 2. Using
this approach we can write wH = , r - , ni |6p>| i)(w6snt (� + 1) n"l + 1
(17.15) and Wq as Wq = |<6sM 5d)PUo(4�2_1)(2� + 3> x (2� - 1)(� +
1)(� + 2) A 2(�2 xV V"* � + (2� + 3)(�2 - �) / r-2 1 A 2" (17.16) In
Eqs. (17.15) and (17.16) the n"summations extend over the continua aswell.
The ranges of n", or equivalently, of energies relative to the 6p and 5d states
of Ba+, which contribute to the sums are different. This point is shown
explicitly by Figs. 17.2 and 17.3. In these figures we show the squared
matrix elements for the initial bound 6sl8h state. To show the matrix elements
between the 18hstate and bound n"t"and continuum e"t" virtual intermediate
states in a consistent waywe have multiplied each of the squared bound
matrix elements |</i"�"11/r^118h)|2 by n"3 and assigned it the energy width
Vn"3 centered at an energy l/2n"2below the Ba+5d or 6p state. In this way
the sum over the bound n"t" states becomes an

370 Rydberg atoms integral over energy of the n"�" series, which smoothly
continues above the series limit. As shown by Fig. 17.2, the (w"g|l/r2|18h)|2
matrix elements from n" slightly lower than n" = 18 provide the dominant
contribution to the n"g sum,whilefor the n"\ sum it is the continuum above the
Ba+ 6p limit which provides the largest contribution. In Fig. 17.3 the
contribution to the n"i sum comes from very low lying states, even below the
bound 6sl8h state, while the contributions to then"h sum come from the region
of the Ba+ 5d limit, and the contributions to the n'\ sum come from well
above the 5dlimit. From Figs. 17.2and 17.3 it isapparent that the energy
range of contributing states to the dipole sums is ~�20% of the Ba+ 6s-6p
energy interval, while for the quadrupole sums the range of contributing
energies exceeds the Ba+ 6s-5d energy interval. It is the energy range of the
contributing terms relative to the ionic energy spacings which leads to the
nonadiabatic effects in core polarization. To show this point explicitly we
shall now ignore the obvious energy variation of Figs. 17.2 and 17.3 and
assume that W6sn� - W6^r = ^6s " W6p and that W6sn� - W5drrr = W6s -
W5d. This approximation is poor for Ba, but excellent for He and alkali
atoms. With these approximations we can remove the energy denominators
from the summationsof Eq. (17.15) and reexpress Wd as K6shi6P)i2 r t v ri't-
1 n� (17.17) If we use the fact that the set of ri'i" functions of arbitrary n",
but constant �", constitutes a complete set of radial functions, we can write
n'T 4 ni\ =V [nt \ n"A(riT 4 nt1V n" /n� 1 1 L l 1 (17.18) Using Eqs. (17.17)
and (17.18) we can write the dipole energy, Wd, as 1 J6^p^ nt) (17.19) 2
Similarly, we can write the quadrupole energy Wq as - w6p) nt). (17.20)

Theoretical description of series perturbation and nonadiabaticeffects 371

0.2 r i 0.1 cc UJ z UJ Ba+5d Ba 5d18h , B a 6 s . 0 Ba 6s18h / \ I I I I 0 0.5
1.0 1.5 2.0 <nCH/r3l18h)2 (10~7aQ~6) Fig. 17.3 The squared matrix
elements <nf|l/r3|l8h>2 ( ), <rch|l/r3|18h> ( ), and (wk|l/r3|18h)2 (� � �)
plotted vs energy to showthe energy distribution ofthe squared matrix
elements. Note that most of the (rcf|l/r3|l8h)2 matrix element lies below the
6sl8h state (fromref. 3). �7 n A (17-21)wq |<6s|rf|f 5(W6s id) 2 i W5d)\ 1 4
1 A \ (17.22) In Eqs. (17.20) and (17.22) the equalities are only approximate
since we have not used all of the n'p and n'd states in computing the dipole
and quadrupole polarizabilities. Aside from this approximation, the above
development shows that the method of Van Vleck and Whitelaw2 reduces to
the adiabatic core polarization model1 when the ionic energy separation is
large compared to the energy ranges spanned by the contributing matrix
elements. However, when the ionic energy spacings are relatively small, as
shown by Figs. 17.2 and 17.3, the nonadiabatic effects become apparent.
When they become smaller yet, series pertubations become important. It is
useful to introduce the factors kd and kq to show explicitly the departure
from the adiabatic model.3 Specifically we may write the dipole and
quadrupole energies of Eqs. (17.15) and (17.16) in a form equivalent to the
adiabatic forms. Explicitly 1 nt), (17.23)

372 Rydberg atoms Table 17.1. Values ofkd and kq calculated for the Ba
6sn� states.a kd q n all 18 23 � = 4 0.945 � = 5 0.953 -0.430 -0.355 � =
6 0.965 1.67 1.89 1 = 1 0.975 1.11 1.14 "(fromref. 3) and _ o ^ ^ r q 2 nt).
(17.24) For the adiabatic case kq = kd= 1. The factors kd and kq, defined
implicitly by Eqs. (17.23) and (17.24), are given by *�, / n"r. - i , " ^6s) <
pn,.(_x - W<6sn� ri't - 1 1 2� + l^W6pn,,(+i-W6sn(i and kn = 21 nl
(17.25) W5d - W6s nl nl 10(4r - 1)(2�+ 3) x (21- 1)(�+ 1)(�+ 2) 2(�2 3
X�j n"� - 2 e_2 ~ W 6sne J (17.26) In Table 17.1 we give the values of kd
and kq calculated for the Ba 6sn� states. The values of kd are n
independent, mainly because the Ba+ 6p state lies well

Experimental methods 373 above the 6s state, and even the lowest lying
6pn� state contributing to the kd sum liesfar from the 6sn� states. For kq
there isasmooth but noticeable n dependence since the lowest lying 5dn�"
states, which contribute substantially to kqy lie relatively near the 6sn�
states, and, as a result, the precise energy of the 6sn( Rydberg state
isimportant. It is also interesting to note that kq isnegative for the 6snh states
sincethe dominant contribution is from the 5d4f state which lies below
the6snh statesofn ~ 20. Since itis almost impossible tocalculate kq accurately
for 6sng states using Eq. (17.26), it is not given in Table17.1. As shown by
Table 17.1, the nonadiabatic corrections become smaller, i.e. fcd�> 1 and
kq �> 1 as �increases. This notion can be appreciated qualitatively by
inspecting Figs. 17.2 and 17.3. As �increases the lowest member of the
contributing ri'i" sums of Eqs. (17.15) and (17.16) moves to higher energies
since the lowest value of n" increases with�". As mentioned earlier, Eqs.
(17.20) and (17.22) are approximate since wehave neglected higher Ba+
states in Eqs. (17.15) and (17.16). In a more accurate calculation one would
apply the approach we have outlined to each contributing state of the Ba+
core, astraightforward procedure. Itis also important tonote that the
contributions from higher lyingstates of Ba+ will bemore adiabaticthan those
from lower lying states. Experimental methods Three different methods have
been used to make microwave resonance measurements of intervals between
alkaline earth Rydberg states. In allof these measurements state selective
laser excitation of alkaline earth atoms in a beam was combined with one of
three forms of state selective ionization of thefinalstates. Thefirstdetection
technique, delayedfieldionization, used by Vaidyanathanet al.6 to measure
4snf 1 F3-4sng X G4 intervals in Ca, is illustrated by the timing diagram of
Fig. 17.4. They optically excited a beam of Ca atoms in two steps, 4s�>4p,
4p^>nf. The second step required the presence of a small field of 20 V/cm
during the excitation. About 200 ns after the laser excitation thefieldwas
turned off, and 1 jus afterwards the atoms were exposed to a 1.6 jus pulse of
microwaves. Finally, about 25jus later allthe Rydberg atomspresent
wereionized by afieldionization pulse. Thismethod is based upon the fact that
the lifetimesof the 4snf*F3levels are much shorter than 25jus, while the
lifetimes of the 4sng!G4 states are approximately 25jus. For example, the
lifetimes of the24f and24g states are 2.5(5) and 25 jus respectively.6 Only if
atoms are driven from the ni to the ng state by the microwaves will there be
Rydberg atoms left after 25 jus. The microwaves must be turned off,
otherwise all the atoms will decay from the ni state, albeit with half the usual
rate. The experiment of Vaidyanathan etal. was done at a temperature of 77
K. In similar experiments with Ba and Cs done at

374 LASER STARK PULSE MICROWAVE PULSE FIELD IONIZING

PULSE Rydberg atoms A TIKAC A VIC -*\ r�-2OOnsec 2OV/cmj l^secjk-
l.6/x.sec-* i >> 1 Fig. 17.4 Timing diagram for the Ca4snf ^^-sng
*G4experiment. A Stark pulse is used to allow excitation of the F state. The
microwave pulse drives the F ^ G transition, and the field ionizing pulse is to
detect the G state (from ref. 6). -e o 27690 27695 FREQUENCY (MHz) Fig.
17.5Field ionization signalfor the 4s25f1F3 -> 4s25g!G4transition in Ca
(from ref. 6). room temperature, the black body radiation induced transitions
from the initially populated statesto longer lived statesproduced a significant
background signal.3'7 In Fig. 17.5 we show the 4s25f 1 F3-4s25g *G4
resonance observed by Vaidyanathan et al.6 As shown, it has a good signal to
noise ratio and it is �^1 MHzwide, close to the 0.6 MHz linewidth expected
for the 1.6 JUS microwave pulse. The second technique,
selectivefieldionization, was used by Gentile et al.8 to measure many
intervals between low�states of Ca. In their experiments the laser
excitation, microwave transition, and detection wereseparated spatially
aswell as temporally, an approach similar to the one of Goy etal.9 described
in Chapter 16.

Low angular momentum states and seriesperturbations 375 Finally, a method

peculiar to alkaline earth atoms was used by Cooke and Gallagher.10 They
took advantage of the fact that the optical transitions from the Sr 5snd states
to the autoionizing 5pnd states occur at different wavelengths than do the
transitions from the 5snf states to the 5pnf states. The reason for this
dependence of the spectra on �quantum defect differences isdiscussed in
Chapter 21. In their experiment two pulsed dye lasers were used to excite Sr
atoms in a beam from the ground 5s5s state to the 5snd state via the
intermediate 5s5p state. The Sr 5snd atoms were exposed first to
microwaves for 1 jus and then to a third laser pulse tuned to the 5snf�> 5pnf
transition. After the third laser pulse the atoms were exposed to a small field
pulse which extracted the ions from the autoionizing 5pnf states, but did not
field ionize the 5snd atoms. Using this method they measured the 5s(n+2)d 1
D2�> 5snf *F3 transitions for 36 < n < 38. Low angular momentum states
and series perturbations Gentile et al.s have measured the 26-80 GHz
intervals between the low angular momentum states of Ca. From these
intervals they have determined the quantum defects of the observed series
and fit them to two forms. The first is \ * (17.27) (n - do)z (n - doy which is a
form often used to take into account the slow variation of the quantum defect
with energy. In the alkaline earth series if there isaperturber in the bound
series the energy perturbation of Eq. (17.14) can be represented in the
quantum defect as8 * a (17.28)+ 2 (n - d'oy (n - dby - 2W p where Wp is the
energy of the perturber. The quantum defects of the Ca 4sns ^o, 4snp 1 P1,
4sns3S1 and 4snp3P1 states are all slowly varying enough to be fit to Eq.
(17.27), while the 4snd ! D2 and3D2 states are much better fit by Eq.
(17.28). The values of the parameters extracted from the microwave
measurements are given in Table 17.2. A measure of the extent to which the
quantum defects of the 4snd series are dominated by perturbers isthat, while
the quantum defects of the 4snd2D2states are �1.2, the value of 6$ given in
Table 17.2 is0.88.

376 Rydberg atoms Table 17.2. Quantum defects ofthes.p, and d states ofCa.a
4sns \ d0 = 2.33793016(30) <5i=-0.1143 (30) 4sns % d0 =2.44095574 (30)
d1 = 0.3495(30) 4snp % d0 = 1.88558437 (30) dx = -3.2409 (50) d2 = -23.8
(25) 4snp 3 Pi <50= 1.96470945 (30) dx = 0.2276(30) 4snd *D2 <50' =
0.88585(50) <V = 0.126 (40) a = 9.075 (90) x 10"4 4snd 3 D2 <50' =
0.88334(50) <V = -0.025 (40) a = 8.511 (90) x 10"4 fl (fromref. 8) Using the
Sr 5snd *D2 quantum defects extrapolated from the calculations of
Esherick11 and their observed 5s(n + 2)d 1 D2-5snf X F3 intervals, Cooke
and Gallagher10 determined the 5snf X F3 quantum defects to be 0.0853.
Nonadiabatic core polarization Perhaps the most interesting aspect of
microwave spectroscopy of alkaline earth Rydberg states is that it affords
easy access to the nonadiabatic effects in core polarization, and such
experiments have been done with both Ca6 and Ba.3 Vaidyanathan etal.6
measured the Ca4snf1F3-4sngXG4 intervals given in Table 17.2. They
compared their measurements to adiabatic and nonadiabatic theories of core
polarization in two ways. First, they compared their measured 4snf-4sng
intervals to those calculated12 assuming adiabatic and nonadiabatic core
polarization models. Vaidyanathan and Shorer12 used the nonadiabatic
model of Eissa and Opik13 rather than the one of Van Vleck and Whitelaw.2
Not surprisingly, the departures from the adiabatic model are parametrized
with factors identical to kd and kq, although there are small higher order
corrections as well. For both the adiabatic and nonadiabatic core
polarization models they used the Ca+ dipole and quadrupole
polarizabilities, ad = 89(9)al and aq = 987(90)^0, calculated using
relativistic Hartree-Fock wavefunctions.12 The adiabatic model predicts
4snf-4sng intervals a factor of 2 too high and the nonadiabatic model predicts
intervals 10% lower than the measured frequencies.

Nonadiabatic corepolarization 371 Table 17.3. Ba+ Dipole and quadrupole

polarizabilities, ad and aq, determined from an analysisof measured 6sn�
intervalsand calculated using coulomb wavefunctions measured calculated "
(fromref. 3) a (*o) 125.5 122.6 (10) <k) 2050 2589 (100) The 10%
difference isprobably due to core penetration in the 4snistates, which is not
taken into account, and, according to the theoretical work of Vaidyanathan
and Shorer12 should be of the same approximate size as the discrepancy. In
any case, it is clear that the nonadiabatic core polarization model reproduces
the observed intervals quite well, while the adiabatic model is substantially
in error. The second approach used by Vaidyanathan et al. in analyzing their
data was not to assume anything about the 4snf states. Rather the 4snf
*F3quantum defect, determined from optical measurements, was used with
the measured 4snf-4sng intervals to determine the quantum defects of the
4sngstates. The 4snf quantum defect from the optical measurements of
Borgstrom and Rubbmark14 is 9.61(5) x 10~2. Together with the microwave
resonance intervals of Table 17.3 this value leads to a quantum defect of
3.10(8) x 10~2 for the Ca 4sng states of 23 < n < 25.6 Since the quantum
defects of the 23 < n < 25states are the same, to the experimental accuracy, it
is not possible to determine both the dipole and quadrupole polarizabilities
from the 4sng quantum defects alone. The dominant contribution, however, is
from the dipole polarizability. Accordingly it is reasonable to use the
computed Ca+ quadrupole polarizability as a small correction and extract a
good value of the dipole polarizability from the 4sng quantum defect. Using
the nonadiabatic model this process leads to ad = 87(3)�o> m good
agreement with the calculated value of S9al. Using the adiabatic model the
value of ad extracted is 63al, far from the calculated value. An alternative
method of investigating nonadiabatic effects was used by Gallagher et al.,3
who measured the Ba 6s�g-6snh-6sm-6sftk intervals. Starting from the
optically excited 6sng X G4 states of 18< n < 23 they observed the resonant
microwave transitions to the higher �states. They observed doublets in the
one-, two-, and three-photon transitions to the 6s�� states of�> 4. They
did not assign the doublets but simply used the average of the two observed
lines to determine the energies of the 6sn�, � > 4 states. The doublet
splittings are about two orders of magnitude larger than the He magnetic
intervals, and they decrease rather slowly with �, unlike an exchange effect.

378 Rydberg atoms Since more than one rf interval has been measured, it is,
in principle, possible to determine both the quadrupole and dipole
polarizabilities from the microwave resonance data alone. The procedure is
identical to that used earlier for alkali atoms,15 except that the factors kd and
kq must be introduced. If we define P' = kdRy \ nt) (17.29a) and u n t 4 n(
(17.29b) where Ry isthe Rydberg constant, in cm *, for Ba, then the
polarization energy of the n� state (in cm"1) is given by WpoU=-adP'-
aqP'Q', (17.30) and the interval between the n� and nf states is given by \W
= W , , - W , , aqAP'Qf, (17.31) where AP' = P'n� - P'n� and AF Q' =
P'niQ'n� - P'n�'Q'ni: Dividing Eq. (17.31) by AP' to remove the variation
in the dipole polarization energy, Wd, with n and � yields AW AP'Q' � ^
AF=ad+a V ^ - (1732) If we plot the observed intervals by plotting AW/AP'
vs AP'QVAP' we obtain a linear plot, the intercept giving ad and the slope
aq. In Fig. 17.6we show a plot of AW/AP' vs AP'QVAP'. Although the 6sng-
6srch6sni-6snk intervals were all measured, the 6sng-6snh intervals are not
shown in Fig. 17.6 since the 6sng series is perturbed by the 5d7d state at n =
24. Originally Gallagher et al. removed the perturbation from the
experimental 6srcg-6snh intervals by using an expression similar to Eq.
(17.16), and then removed the approximately corresponding term from the kq
sum. Specifically the term from the 5dnd state just below the Ba+ 6s limit
was removed from the kq summation of Eq. (17.26). The fact that the data
modified in this way appeared reasonable in a plot such as Fig. 17.6 was
probably fortuitous. From the 6snh-6sm and 6sm-6snk intervals shown in Fig.
17.6 reasonable values of ad and aqmay be obtained, as shown by Table
17.3. The experimental uncertainties given in Table 17.3 are those from
fitting the data to Eq. (17.32). We also give in Table 17.3 the values
calculated using coulomb wave functions for

Nonadiabatic corepolarization 379 140 135 130 120 115 I I I I I I I I I I I I I I

I I I I I I I -0.006 -0.004 -0.002 0 0.002 0.004 0.006 AP'QVAP' Fig. 17.6 A
plot of the measured Ba A� intervals using the factors kd andkq to correct
for the nonadiabatic effects (from ref. 3). Ba+.16 Incalculating the dipole and
quadrupole polarizabilities onlythe 6p and 5d states were used, leading to
values of ad and aqwhich are too lowby an estimated 1% and 10%
respectively. The agreement of the measured and calculated dipole
polarizabilities is good, but between the measured and calculated quadrupole
polarizabilities it isat best fair. Either the calculation of the polarizability
iswrong or the kq correction factors are wrong. In fact, it is almost certain
that thekq factors ofthe 6snhstates are incorrect due to our having ignored the
spins ofthe electrons. Forthe6sngand6snh states this omission leads to
significant errors. The reason is shown in Fig. 17.3, the plot of the squared
matrix elements contributing to kqfor the 6sl8h state. As shown, the spinless
5d4f state lying1750 cm"1 below the Ba+ 6s limit has a large matrix element,
and, due to its proximity to the bound 6snh states, it makes the largest
contribution to kq. Consequently, the precise location of this state isimportant
to the calculation of kq. It isnot well known for two reasons. First, in reality,
the Ba+ 5d state issplit into the 5d3/2 and 5d5/2 states, which are800 cm"1
apart, and since there are4f5/2 and 4f7/2 states converging to both of the Ba+
5d limits,it isreasonable to expect the real Ba 5d4f states to fall in a range
�800 cm"1 wide around theenergy of the spinless 5d4f state. An 800 cm"1
range is hardly small compared to the 1500 cm"1 energy difference between
the 6sl8h state and the spinless 5d4f state and ispotentially a significant
source of error. Second, the quantum defects of the 5dnf statesare not
zero,17but �0.05.For both of the above reasonswe cannot expect
thevaluesofkq calculated for the 6snf states using spinless wavefunctions to
be particularly accurate. For the 6smandhigher angular momentum
statestheterms contributing to kq all lie �1000cm"1 above the 6s limit and
are thus less susceptible toerrors

380 Rydberg atoms than the 6snh states. For the 6sng states, on the other
hand, kq can be calculated with even less confidence than for the 6snh states.
While the agreement of the measured and calculated Ba+ quadrupole
polarizabilities is not very good, compared to an analysis based on adiabatic
core polarization the agreement in Table 17.3 is superb. The adiabatic core
polarization model leads to ad = 146al and aq = -5800^. The ground state of
Ba+ cannot have anegative quadrupole polarizability. Taken together, the Ca
and Ba experiments show clearly that the nonadiabatic effects in core
polarization in the alkaline earth atoms are important and may be calculated
with some accuracy. References 1. J. E. Mayer and M. G. Mayer, Phys.Rev.
43,605 (1933). 2. J. H. Van Vleck and N. G. Whitelaw, Phys.Rev. 44, 551
(1933). 3. T. F. Gallagher, R. Kachru, and N. H. Tran, Phys. Rev. A 26, 2611
(1982). 4. H. A. Bethe and E. E. Salpeter, Quantum Mechanics of One- and
Two-Electron Atoms (Springer, Berlin, 1957). 5. A. R. Edmonds, Angular
Momentum in Quantum Mechanics (Princeton University Press, Princeton,
1960). 6. A. G. Vaidyanathan, W. P. Spencer, J. R. Rubbmark, H. Kuiper, C.
Fabre, D. Kleppner, and T. W. Ducas, Phys.Rev. A 26, 3346 (1982). 7. K. A.
Safinya, T. F. Gallagher, and W. Sandner, Phys.Rev. A 22, 6 (1980). 8. T. R.
Gentile, B. J. Hughey, D. Kleppner, and T. W. Ducas, Phys.Rev.A 42, 440
(1990). 9. P. Goy, C. Fabre, M. Gross, and S. Haroche, /. Phys. �13,L83
(1980). 10. W. E. Cooke and T. F. Gallagher, Opt. Lett. 4, 173 (1979). 11. P.
Esherick, Phys. Rev. A15, 1920 (1977). 12. A. G. Vaidyanathan and P.
Shorer, Phys.Rev. A 25,3108 (1982). 13. H. Eissa and U. Opik, Proc.
Phys.Soc. London 92, 556(1960). 14. S. A. Borgstrom and J. R. Rubbmark,
J. Phys. B 10, 18 (1977). 15. B. Edlen, in Handbuch der Physik, ed. S.
Flugge (Springer-Verlag, Berlin, 1964). 16. M. L. Zimmerman, M. G.
Littman, M. M. Kash, and D. Kleppner, Phys.Rev.A 20, 2251 (1979). 17. E.
A. J. M. Bente and W. Hogervorst, /. Phys.B 24, 3565(1989).

18 Bound He Rydberg states The He atom differs from the atoms we have
discussed in the last two chapters in that many of the intervals between high
lyingstates can be calculated with ease to an accuracy better than 1%.1>2
Consequently, the challenge is to calculate the intervals more accurately and
to make measurements to test these sophisticated calculations. While the
sophistication of the calculations is substantially beyond the scope of this
book, the measurements of the intervals between the He lsn� states
constitute a significant fraction of the measurements of Rydberg energy level
spacings. Accordingly, the focus of this chapter is on the experimental
measurements which have been made. Theoretical description The
Hamiltonian for two spinless electrons at points r1 and r2 relative to a He2+
core is given by3 where r12 = |r2�r11, and rt and r2 are the positions of the
twoelectrons relative to the He2+ ion. To describe an atom with one electron
in a Rydberg state a reasonable approach, termed Heisenberg's method
byBethe and Salpeter,3 is to separate the Hamiltonian of Eq (18.1) into two
parts. Explicitly, H=HO +HU (18.2a) where �0=-d+l+i+i) (18.2b) \2 2 rx
r2f and #! = � + �. (18.2c) r2 rl2 Using the Hamiltonian Ho of Eq. (18.2b)
in the time independent Schroedinger 381

382 Rydberg atoms equation the problem is separable into independent

solutions for the two electrons, and the solutions can bewritten as (18.3)
where <pnieimi and ipm(imi are the wavefunctions for He+ and H.
Correspondingly, the energies aregiven by 2 1 relative tothe double
ionization limit. We are, atthe moment, only interested in the bound Rydberg
states of He inwhich electron 1 is in the Isstate, so we shall specify the
wavefunction ofEq. (18.3) as For consistency with other bound Rydberg
states we shall measure the energies relative tothe ground state ofHe+.
Expressed in this way, the energy ofthe state given inEq. (18.5)is Wn�m =-
l/2n2. (18.6) Following the Pauli exclusion principle we must
antisymmetrize the wavefunctions when weinclude the spins of the two
electrons. The wavefunction isgiven by a product ofspatial and spin wave
functions, i.e. 1 V�nem-V2 is *i tpn�m r2 _ l5 r2 VWmri � . ^ ^ ^ Here
0� is the spin wavefunction, and the + and � signs refer to the orthogonal
singlet, S = 0and triplet, S = 1 states, where S isthe total electron spin. Using
the wavefunctions of Eq. (18.7) we cancalculate the first order corrections to
Eq. (18.6) using the Hamiltonian ofEq. (18.2c). Explicitly, '�^mdr!dr2.
(18.8) Assuming that 0l s andVWm a r e orthogonal we can express the
integrals implied by Eq. (18.8) in terms of the direct and exchange
integrals,/and K. The/integral compensates for theincomplete screening of the
outer electron from the doubly charged core. Equivalently, the / integral
reflects the penetration ofthe He+ core by the outer nl electron. From either
point ofview, itis apparent that / < 0. The exchange integral Kcan be viewed
asthe rate at which the two electrons exchange their states. Explicitly, we can
express thepenetration and exchange energies as Wpen andWexas Wpeo= / =
I �i (ri) |2I un(m (r2)|2 dTi dr2 (18.9) 1 Vu r2} and
Theoretical description 383 Wex = K=\ u\{rl)u^m{rl)u1{Y2)unejY2) � dr x
dr2. J rn (18.10) Since both / and K integrals depend upon the spatial
overlap between the ground state He+ wavefunction andthe Rydberg nlm
wavefunction, they decrease very rapidly with increasing
�andcanbeexpressed asrapidly converging series.3 In addition
tothefirstorder corrections to theenergy given by Eq. (18.11), we must
account for the polarization of the He+ core by the field from the outer n�
electron.4'5 As we saw in Chapter 17, this correction is really a second
order correction,6 but ifwritten in terms of the polarizabilities of He+
itappears as a first order correction. Core polarization is relatively most
important in�> 2 Rydberg states, for which the inner classical turning point
occurs at r2 > 6%, while the classical outer turning point ofthe He+ Is
electron is at la0. Thus, it isan adequate approximation to assume that r2 >
r1? ignore exchange effects, and use the unsymmetrized wavefunction of Eq.
(18.5) to calculate the polarization energy. In this approximation
thepolarization energy, Wpol, isgiven by4'5 W � �� (r ~ 4 \ _ ^ � / V ~ 6
\ (\9K\W pol � \ 2 Intm \ 2 /n�m ? ^J-O.J-J-y where (r'k)n^m is the
expectation value of r"k for a hydrogenic nlm state. Explicit expressions for
these expectation values are given in Chapter 2, and the He+ polarizabilities
are given by ad = 9ao3/32 and aq =15ao5/64.5 When the exchange,
penetration, and polarization energies are included, the energies of the singlet
andtriplet n�m states are given by W�n�m =^-2 + Wpen � Wex. (18.12)
2n The penetration and exchange energies are typically of the same size and
decrease rapidly with increasing �.3 The polarization energy decreases
more slowly with �, andfor� > 2it isthe dominant contribution to the
deviation from the hydrogenic energy in Eq. (18.12). For � = 2 thethree
energies are comparable, and for �< 2the exchange and penetration energies
exceed the polarization energies. Wpo{ and Wpen are both negative, and
Wexis positive. Introducing spin adds significant magnetic interactions,
which may be accounted forwith the Breit-Bethe theory.3 The theory is based
on the unsymmetrized wavefunctions used to calculate the polarization
energy. There are two terms which make significant contributions, the spin-
orbit interaction and the spin-spin interactions. These operators have the
matrix elements3 <Vn<sj\HSo\y>n�S'j) = ^ s o (USS'J) (18.13) and (xjj H
\\l) ' ) � AM (�SSr /) (18 14) where 2
384 Rydberg atoms Table 18.1. Matrix elements of the Breit-Bethe spinorbit
andspin-spin operators, Eqs. (18.13) and (18.14). S,S',J Mso(tSS'J)a>b
Mss(�, SS'J)a 0,0, � 0 0 1, 1, � + 1 -� 2�/(2� + 3) 1,1, � - 1 -2 1, 1,
�- 1 � + 1 (2�+2)/(2� - 1) 1,0, � -3V�(� + 1) 0 a (S = S' matrix
elements from ref. 3) (see ref. 10) b (S = 0, S' = 1element from ref. 7) J =
�+ S, and a is thefinestructure constant, 1/137.07. Thevaluesof Mso(tSS'J)
and MSS(�SS'J) are given in Table 18.1. The diagonal elements, S = Sf = 1,
are given by Bethe and Salpeter,3 and the off diagonal element is given by
Milleret al.7 Due to the (r^3) dependence, the magnetic fine structure
decreases slowly with increasing �. For low � the polarization and
exchange energies are both far larger than the magnetic fine structure
energies. The singlet states are well above the triplet states, and in computing
the magnetic fine structure the off diagonal spin orbit matrix element of Eq.
(17.13) can beignored. On the other hand, for high � states the polarization
energy is largest, followed by the magnetic fine structure energies, followed
by the penetration and exchange energies. In this case the states are no longer
all good singlets and triplets. The/ = � �1 states are triplets, but theJ =(
states are roughly 50-50mixtures of singlet and triplet. In the limit of the zero
exchange energy the energies of the two / = �states are symmetrically
displaced from the polarization energy by � AMSO(�, 1, 0, �).
Experimental approaches Most of the high precision spectroscopy of He
Rydberg states has been done by microwave resonance, which is probably
the best way of obtaining the zero field energies. Wingetal.8^12used a 30-
1000juA/cm2 electron beam tobombard Hegas at 10~5-10~2 Torr. As
electron bombardment favors the production of low � states,itispossible to
detect A�transitions driven bymicrowaves. The microwave power was
square wave modulated at �40Hz, and the optical emission from a specific
Rydberg statewas monitored. When microwave transitions occurred toor

Experimental approaches 385 Fig. 18.1The resonance module of MacAdam

and Wing: 1, heater; 2, dispenser cathode; 3, accelerating
grid;4,grid;5,grid;6,the "cage,"consisting ofwires joiningthe perimeters of
grids5 and 6; 7, grid; 8,collector; 9. and 10, X-band waveguide horns.The
grid supports are nonmagnetic stainless-steel plates, 2.54 cm square. The
cage is2.54 cmlong by 1/6cm diameter (from ref. 11). from the state
monitored, an increase or decrease in the detected emission was observed in
phase with the modulation of the microwave power. In Fig. 18.1we show the
apparatus used by MacAdam and Wing.11 Electrons from a dispenser
cathode were accelerated and passed down the axis of a cage formed by W
wires, where they excited He atoms to Rydberg states. The cage, which
wasperiodically heated to incandescence, ensured that there were no stray
electricfieldsdue to long term buildup of oil or other charge trapping
insulators near the Rydberg atoms. The cage was positioned between two X
band microwave horns, which brought microwave radiation to the Rydberg
atoms. Microwave power was fed into the cage from one or both horns
allowing the simultaneous use of two frequencies. Light from the Rydberg
atomswas collected by an ellipsoidal light pipe with one focus at the center
of the cage of Fig. 18.1 and the other at the entrance slit of a 1/4 m
monochromator. A typical resonance signal is shown in Fig. 18.2, for the two
photon 9d X D2 �> 9d ^4, and

386 Rydberg atoms 17650 17700 17750 FREQUENCY (MHz) Fig. 18.2
Chart recorder trace of the 91D2-91'3G4 resonances. The resonances are
power broadened by a factor of 2-3 for purposes of display. Sweep rate 0.4
MHz/s, lock-in time constant 2s (from ref. 11). 9d ! D2 �� 9d3G4
transitions.11 The resonances of Fig. 18.2 were obtained by monitoring the
9d 1 D2-2p 1 F1 fluorescence at 3872 A. The transitions shown in Fig. 18.2
are power broadened by a factor of 2-3 above the instrumental linewidth.
They are also exhibit an ac Stark shift, linear in the microwave power, and to
extract the correct intervals the measurements must be made at several
powers and the measured resonance frequencies extrapolated to zero power.
For reference, the zero power intervals corresponding to Fig. 18.2 are 1 D2-
1G4, 17697.065(41) MHz, and 1 D2-3G4, 17661.096(22) MHz. Perhaps the
most precise measurements are those made using fast He beams. The first
experiments were those of Cok and Lundeen.13 A duoplasmatron ion source
produced a 5-20 keV beam of He+, usually with an energy of 13keV. The
beam passed through a charge exchange cell containing Ar at a pressure of
100 juTorr. Unlike electron bombardment, charge exchange produces useful
populations in higher �states. For example, Cok and Lundeen estimated that
there was a flux of 108He 8f atoms/s in their beam. The He beam passed
coaxially down a piece of circular waveguide. The waveguide had
microwaves travelling either
Experimental approaches 387 C02 Loser RF Pre-Pump Post Pump
Deflection /C02 Laser Defle Shield \ ( Detection / Channeltron Magnet t \ V /
J C \ VaradayCup � / Stark Ionizer r o M f h Collimator Ion Source ^e n |
Spectroscopy Collimator Fig. 18.3Fast He beam apparatus used to measure
theradio frequency intervals between the n=10 g, h, iandkintervals (from ref.
14). parallel or antiparallel to the He beam, and by making measurements
with the microwaves travelling in both directions it was possible to cancel
the rather large Doppler shifts, 0.1% of the microwave frequency.
Downstream from the waveguide either the 4d 1 D2-2p 1 P1, or 4d 3 D2-2p
3 PX emission was detected, depending upon whether X F or 3 F Rydberg
states were being studied. In either case, it was the second step of the 8f �>
4d �� 2p cascade. There was minimal emission from the 4d state atoms
populated directly bycharge exchange since they decayed before reaching the
detection region. Using this technique Cok and Lundeen measured n = 1 and 8
f-g, f-h, and f-i intervals. The minimum resonance linewidth was limited to
about 4 MHz by the transit time of atoms through the waveguide.13 Palfrey
and Lundeen modified the previous approach by using a different detection
scheme.14 They used a Doppler tuned CO2 laser to drive transitions from
specific �states of n = 10to n = 27and subsequentlyfield ionizing the ft = 27
He atoms. A diagram of their apparatus is shown in Fig. 18.3.The fast He
beam initially contains all 10�states, but one of the �>4 states isdepleted
by driving the transition to the ft=27 state with the CO2laser. The depleted
beam passes through the resonance region where a rf field, when tuned to a
resonance, drives the 10 (� + 1)�>� 10� transition, repopulating the
empty 10� state. The increase in population of the 10� state at the
10�-10(�+l) resonance is detected by using the CO2 laser beam a second
time to drive the 10� atoms to the n = 27 state andfield ionizing them. This
approach ismuch more sensitive to transitions between high � states than
one based on optical detection, and they used it to make very precise
measurements of lOg-lOh-lOi-lOk intervals.14 This approach has been
continuously refined, and, using it, Hessels et al. have been able to measure
transition frequencies to better than 1kHz.15 A completely different approach
is anticrossing spectroscopy,16 employed by Miller etal ,717Beyer and
Kollath,18"21 and Derouard etal.22'23 This method is, in principle, a direct
way of measuring the mixing of the singlet and triplet states and the singlet-
triplet x Lr^>Lt separation and has been primarily applied to theXD2 and 3
D2 states. To understand the method we begin with the magnetic field

388 Rydberg atoms B(KG) Fig. 18.4. Energy level diagram (top) of the n
=613D states asa function of magnetic field and actual spectra (bottom) of the
n = 613D anticrossing. Each of the spectral traces was obtained in
approximately 10 min running time. Top trace shows the decrease in light
intensity of the 6*D emission line at 4144 A, while the lower trace showsthe
corresponding increase in the 63D emission at 3819A (from ref. 17).
dependence of the energies. In Fig. 18.4, we showaplot of theHe6d
energylevels as a function of magnetic field.17 To the zerofieldHamiltonian
we add where juB is the Bohr magneton and g�and gs are the g factors of
the orbital and spin magnetic moments, 1 and 2, respectively. It is useful to
define the average energy of the singlet and triplet states, Wav, which is
given by W�v =-^+Wp e n + Wpol. (18.17) The spin-orbit energies are very
small compared to the other energies, so we can treat the spin and orbital
angular momenta as being uncoupled. If we ignore

Experimental approaches 389 the offdiagonal spin-orbit matrix element of Eq

(18.13), an excellent approximation for the He nd states, the energies W+ and
W_ of the *D2 and 3 D2 states, respectively, are given by W+ - Wav = Wex
+ fiBg�m�+ (18.18a) W_ - Wav = - Wex -A+ vB(gsms_ + gem�_)
(18.18b) where m�+ is the ra�value of the *D2 state and ms_ and m�_
are the ms and ra� values of the 3 D2 state. In the magnetic field the total
azimuthal angular momentum, ray = ms + m�, is a good quantum number,
and, as shown by Fig. 18.4, 6d 1 '3D2 states of the same m;, and m�_ =
ra�+ � 1intersect at a field of 15 kG.17 Equating the X D2 and 3 D2
energies of Eq. (18.18) for m�+ = m�.+l and ms_ = 1yields W+ - W_ =
/*5C, (18.19) where #c is the magnetic field at the intersection of the singlet
and triplet levels. If there were no coupling between the singlet and triplet
levels, theywould cross at Bc. However, the off diagonal spin-orbit matrix
element of Eq. (18.13), which weignored in computing the energies of Eq.
(18.18), couples them, and there isan avoided crossing, or anticrossing, at
Bc. At the anticrossing the two levels are separated by twice the off diagonal
matrix element of Eq (18.13), i.e. by 6AV5, and the eigenstates are not the
*D2and 3 D2 states but 50-50 mixtures of the two states. In the experiment,
electron beam excitation from the ground 1 S0state produces primarily
singlet Rydberg states. If the 6d *D2-2p 1 P1 fluorescence at 4144 A is
observed, as the magnetic field is swept through the anticrossing a drop in
the fluorescence is observed at the anticrossing, as shown in Fig. 18.4.17The
origin of the decrease in the fluorescence can be understood in the following
way. The same number of Rydberg atoms isexcited bythe electron beam
irrespective of the value of the magnetic field. Away from the avoided
crossing only the *D2 state is excited, but at the avoided crossing half of the
excitation is into each of the two mixed states. Away from the avoided
crossing the X D2 atoms excited by the electron beam can only decay to
lower lying singlet states, but at the avoided crossing the two mixed states
can decay to lower lying triplet states aswell. To the extent that decay to the
lower lying triplet state occurs, the singlet fluorescence is diminished. If the
decay rates to lower lying singlet and triplet states are equal, at the
anticrossing the singlet emission drops by 50%. On the other hand, if the
6d3D2-3P1 radiation at 3819 A is observed, normally very little light is
detected, and a sharp increase is seen at the anticrossing, as shown in Fig.
18.4. According to Eq (18.10) the field of the anticrossing gives the singlet
triplet splitting. The width of the anticrossing gives the spin-orbit matrix
element. In practice there are several factors which complicate anticrossing
spectroscopy. The four anticrossings of Fig 18.4 apparently occur at the same
field, given

390 Rydberg atoms QUARTZ LIGHT PIF ANODE GRID 'CATHODE

HEATER QUARTZ SUPPORT Fig.18.5. Apparatususedtoobserve
anticrossingemissionfromthe7*D state (fromref. 7). by Eq. (18.19). However,
we have ignored corrections such as the quadratic Zeeman effect and the
motional Stark effect, which split the four anticrossings, but not enough to be
well resolved.5 Unfortunately, their relative strengths are not well known
since the populations produced by electron impact excitation are not
known.17 In most of the He 1 D2-3D2 anticrossing measurements the fact
that there are four overlapping anticrossing signalsleads tounavoidable
ambiguitiesin the interpretation of the observed signals. Beyer and Kollath20
removed someof the ambiguities by applying an electricfieldperpendicular to
the magneticfieldto split the unresolved anticrossings into well resolved
anticrossings. Using this technique they were able to make the most precise
anticrossing measurements of 1 D2-3D2 splittings. Applying electric fields
also enabled them to observe anticrossings of the nd states with states of
higher �,and in thisway they made the first determinations of the energies of
the � > 3states.21 A typical experimental arrangement for anticrossing
spectroscopy is shown in Fig. 18.5.7 A Cu vacuum chamber,
whichfitsbetween the pole faces of amagnet, encloses the assembly of Fig.
18.5 and contains flowing He at a pressure of 1 mTorr. The cathode produces
�100 juA of electrons which are accelerated by applying 30-50 V between
the cathode and grid and a comparable voltage between the grid and the
anode. The light, leaving in the direction perpendicular to
thefield,iscollected by alight pipe, and the desired wavelength selected byan
interference filter before the light is detected by a photomultiplier tube. A
final technique which has been used by Panock et al24'25 is laser magnetic
resonance. They excited He atoms in a 20-140 kG magnetic field by electron
impact. They used alinetunable CO2laser to drive atoms in the 7s1S0state to
the

Quantum defects and fine structure 391 n = 9 � > 2 levels, which contained
9p1P1 character due to the strong magnetic field. By changing lines, the laser
frequency can be varied in steps of �1.3cm"1, and the magnetic field is
continuously tunable. Together these two tuning methods give continuous
tuning over a broad frequency range. The 7s ��9� transitions were
observed by monitoring the 7s 1 S0-2p 1 P1 radiation as the magnetic field
was slowly scanned through the resonances. To observe a 7s�> 9�
transition requires that there be a 9p admixture in the 9� state. For odd �
this admixture is provided by the diamagnetic interaction alone, which
couples states of �and ��2, as described in Chapter9. For even �states
the diamagnetic coupling spreads the 9pstate to all the odd 9� states and the
motional Stark effect mixes states of even and odd (. Due to the random
velocities of the He atoms, the motional Stark effect and the Doppler effect
also broaden the transitions. Together these two effects produce asymmetric
lines for the transitions to the odd 9� states, and double peaked lines for the
transitions to even 9� states. The difference between the lineshapes of
transitions to the even and odd 9� states comes from the fact that the
motional Stark shift enters the transitions to the odd 9�states once, in the
frequency shift. However, it enters the transitions to the even 9� states
twice, once in the frequency shift and once in the transition matrix element.
Although peculiar, the line shapes of the observed transitions can be analyzed
well enough to determine the energies of the 91 states of �>2 quite
accurately.25 Quantum defects and fine structure The quantum defects of the n
= 10 states of He are given in Table 18.2. The quantum defects for the s and p
states are derived from optical measurements, specifically from the term
energies given by Martin.26 The quantum defects of the � = 8and9 (10�
and 10m) states are calculated using Eq. (18.2), and the quantum defects of
the 2 < � < 7 states are obtained from the calculated � = 8 quantum defect
and the measured intervals reported by Hessels et al.15 The quantum defects
given by Table 18.2 are by no means indicative of the possible precision of
the measurements. For example, Hessels etal. have reported measurements of
n = 10intervals which have precisions of parts per million.15 One of the
more interesting aspects of the He atom is the evolution of the / = � states
from well defined singlet and triplet states at low � to states which are
mixtures of singlet and triplet at high �. This evolution isshown inFig. 18.6
for the 9d-9g levels. The intervals are taken from Farley et al.12 In the 9d
state the exchange splitting is orders of magnitude larger than the spin-orbit
and spin-spin splittings, and the 9d states are good singlet and triplet states.
The amplitude of X D2 wavefunction in the 3 D2 states is ~10~2 and vice
versa.9 In the 9f state the exchange energy is the same order of magnitude as
the magnetic energies. As a

392 Rydberg atoms Table 18.2. Quantum defects of the He states of n ~ 10 3o

a ^1 1T> a M 3T> a "l *D2* 3 D* ib hb ib kc emc 0.14 0.30 -0.012 0.063
2.08 X 2.82 X 4.26 X 1.17 X 4.29 X 1.90 X 7.79 X 3.68 X 1.93 X 10"3
10~3 lO"4 10"4 10"5 lO"5 10"6 10"6 10~6 fl (fromref. 25) 6 (fromref. 15) c
(fromEq. (18.11) 1 D2 (a) 6662.00 (b) 3D, 71.06 58.12 f 3 D 3.79 U 3 Fig.
18.6.Exchange and magnetic energies of (a) the He 9dstates, (b) the He
9fstates, and (c)the He 9g states. In (a)the9d states,the singlettriplet
splittingis not showntoscale.The 9d states are good singlet and triplet states.
The 9f states are approximately singlets and triplets, but the3F3 state isbelow
the3F4 state. The 9gstates do not display singlet-triplet structure. All
intervals are given in MHz.

References 393 result, while the *F3 state is still clearly removed from the
triplet states, the off diagonal spin-orbit matrix element depresses the energy
of the3F3 level to below that of the3F4 level. Correspondingly, the ni1?3F3
states are not good singlets and triplets; there is a singlet amplitude of 0.55 in
the triplet state and vice versa.10 In the 9g state the exchange energy is
negligible, and only the / = 3 and / = 5 states are good triplets. The/ = 4
states are 50-50 mixtures of singlet and triplet. For the 9g state the intervals
shown in Fig. 18.6can be calculated within 1 MHz usingEqs. (18.4) and
(18.5). However, for the 9d and 9f statesa more sophisticated approach is
required. Using many body perturbation theory Chang and Poe27 have
calculated 1 D2-3D2 and ^ - ^ intervals of 6615 and 69.09 MHz, in good
agreement with the experimental results shown in Fig. 18.6. However, using
an extended adiabatic model Cok and Lundeen28 were able to reproduce all
six 9d and 9f intervals to within less than 1MHz. References 1. RJ.
Drachman, Phys.Rev. A 31, 1253 (1985). 2. G.W.F. Drake, Phys.Rev. Lett.
65,2769 (1990). 3. H.A.Bethe and E.E. Salpeter, Quantum Mechanics of
One- and Two-Electron Atoms (Springer, Berlin, 1957). 4. J.E.Mayer
andM.G. Mayer, Phys.Rev. 43, 605(1933). 5. C.Deutsch, Phys.Rev. A 2, 43
(1970). 6. J.H.van Vleck and N.G. Whitelaw, Phys.Rev. 44, 551 (1933). 7.
T.A.Miller, R.S. Freund, F. Tsai, T.J. Cook, and B.R. Zegarski, Phys. Rev.A
9,2474 (1974). 8. W.H. Wing andK.B. MacAdam, inProgress in Atomic
Spectroscopy, eds. W. Hanle andH. Kleinpoppen (Plenum, New York, 1978).
9. K.B. MacAdam and W.H. Wing, Phys.Rev. A 12, 1464 (1975). 10. K.B.
MacAdam and WH. Wing, Phys.Rev. A 13, 2163 (1976). 11. K.B. MacAdam
and W.H. Wing, Phys.Rev. A 15,678 (1977). 12. J.W. Farley, K.B.
MacAdam, and W.H. Wing, Phys.Rev. A 20,1754 (1979). 13. D.R. Cok
andS.R. Lundeen, Phys.Rev. A 23, 2488 (1981). 14. S.L.Palfrey andS.R.
Lundeen, Phys.Rev. Lett. 53, 1141 (1984). 15. E.A.Hessels, P.W. Arcuni, F.J.
Deck, andS.R. Lundeen, Phys. Rev. A 46, 2622 (1992). 16. T.G.Eck, L.L.
Foldy, and H. Weider, Phys. Rev. Lett 10, 239 (1963). 17. T.A.Miller, R.S.
Freund, andB.R. Zegarski, Phys. Rev.A 11, 753 (1975). 18. H.J.Beyer
andK.J. Kollath, /. Phys. B8, L326(1975). 19. H.J.Beyer andK.J. Kollath, /.
Phys. B9, L185(1976). 20. H.J.Beyer andK.J. Kollath, /. Phys.B 10,
L5(1977). 21. H.J.Beyer andK.J. Kollath, /. Phys.B 11, 979 (1978). 22. J.
Derouard, R. Jost, M. Lombardi, T.A. Miller, andR.S. Freund, Phys. Rev. A
14, 1025 (1976).

394 Rydberg atoms 23. J. Derouard, M. Lombardi, and R. Jost, / . Phys.

(Paris) 41, 819 (1980). 24. R. Panock, M. Rosenbluth, B. Lax, and T.A.
Miller, Phys.Rev. A 22,1050 (1980). 25. R. Panock, M. Rosenbluth, B. Lax,
and T.A. Miller, Phys.Rev. A 22,1041 (1980). 26. W.C. Martin, /. Phys.
Chem.Ref. Data 2, 257 (1973). 27. T.N. Chang and R.T. Poe, Phys.Rev. A
10, 1981 (1974). 28. D.R. Cok and S.R. Lundeen, Phys.Rev. A 19,1830
(1979).

19 Autoionizing Rydberg states Autoionizing states are those atomic states in

which there are at least two excited electrons which together have enough
energy that one can escape from the atom. We shall consider only states in
which there are two excited electrons, one of which is in a Rydberg state.1
From a spectroscopic point of view an autoionizing state isone which
iscoupled to acontinuum, and from acollision point of viewitis a long lived
scattering resonance. In other words, autoionization is located at the
intersection of collision physics and spectroscopy, and the theory commonly
used to describe autoionizing states, quantum defect theory, is based on
scattering theory. Basic notions of autoionizing Rydberg states
Itisstraightforward to extend the picture of the spinless Ba atom given in
Chapter 17 to autoionizing states. We again use the Hamiltonian of Eq. (17-
1), which separates into H = H0 + Hl9 (19.1) where [y y ^] (19-2) and H1 =
-/(r2) + i + � � (19.3) Here rx and r2 are the locations of the two electrons
relative to the Ba2+ ion, and r i2 = lri ~~ r i\- I" Eqs. (19.2) and (19.3) f(r)
is the potential experienced by an electron at a distance r from the Ba2+
core. As r�� ��,f(r) �> 2/r. Using Ho only, the Schroedinger equation
H0*(r1,r2) = W*(r1,r2) (19.4) is separable into the two equations 395

396 Rydberg atoms rv? l � h/(rx) 0(rx) = W^rj), (19.5a) L2 J and � I�

H�I V'fe) = W2t/^ (r2), (19.5b) \2 r2/ in which (19.6a) and W= Wx + W2.
(19.6b) Here 0 is a Ba+ wavefunction and xp is a hydrogenic wavefunction.
The wavefunction of Eq. (19.6a) isagood zero order representation of the Ba
atom as long as the inner turning point of the outer ni electron is at a larger
radiusthan the outer turning point of the Ba+ n'V electron. Since the two
wavefunctions are independent, the energies do not depend on m', m, or
�,and wecan write the energy of the n'l'nt state relative to the ground state of
Ba2+ as WW�� = W W - - ^ - (19.7) In Physically, the wavefunction of
Eq. (19.6a) corresponds to adding one electron to Ba2+ to make the Ba+
n'�f states and then adding the second n� electron to form the neutral Ba
n'Vni states. On each state of Ba+ there is built an entire system of Rydberg
states and continua, as shown in Fig. 19.1 for the three lowest lying states of
Ba+. Note that the Rydberg states converging to the 5d and 6p states of Ba+
are above the Ba+ 6s state. With the independent electron Hamiltonian Ho
these states can only decay radiatively. They are not coupled to the
degenerate continua above the Ba+ 6s state. Adding H1 to the Hamiltonian
introduces the couplings shown bythe arrows in Fig. 19.1. First, and most
important, it couplesthe doubly excited states above the lowest limit to the
degenerate continua, introducing autoionization. Second, it leads to couplings
between nominally bound series converging to different Ba+ limits. The
perturbations of the regularities of the energies of the bound Rydberg series
are the result of this interaction between nearly degenerate states
convergingto different limits,whilecorepolarization is the result of the
coupling between energetically well separated states. These two phenomena
were discussed in Chapter 17. One of the most interesting examples of
interseries interaction isthat between autoionizing Rydberg series. It is often
the case that two interacting states overlap, due to their finite autoionization
widths, leading to striking spectral profiles. We shall, in the following
chapter, discuss these states in detail,

Basic notions of autoionizing Rydberg states 397 Fig. 19.1 The three lowest
lyingenergy levelsof Ba+; 6s, 6p, and 5d ( ); and the neutral Ba Rydberg
energy levels ( ) and continua (///) obtained by adding the second valence
electron to these Ba+ levels.The horizontal arrows showthe possible
interactions between channels associated with different ion levels.
Interactions with other bound states lead to series perturbations while
interactions with continua lead to autoionization. but for the moment we focus
on isolated autoionizing states coupled only to the degenerate continua,
ignoring any interseries interaction between autoionizing series converging to
different limits. Finally, although it isnot shown in Fig. 19.1, the quadrupole
term of Ht lifts the degeneracy of n'Vni states of �', t > 1; i.e. states of
different total angular momentum do not havethesame energy duetothe
differing orientations of �' relative to t. The autoionization rate F of a Ba
6pn� state converging to the Ba+ 6p state is given by the product of the
squared coupling matrix element and the density of final continuum states.
Explicitly,2 T = 2jt( Y K6pn�|//1|6s��')|2+ Y, l<6p^|//1|5de�")|2j, (19.8)
�'=��1 �"=��1 where 6ss�' and5ds�"arethe continua associated
with the 6s and5d statesof Ba+. To account properly for the density
offinalstatesEq. (19.8) requires that the 6s��' and 5ds�"wavefunctions be
normalized per unit energy. A 6pra�state with an autoionization rate F is
broadened, and its spectral density is described by a Lorentzian of width F
centered at its energy Wo. A reasonable first approximation to the
wavefunction isobtained by multiplying the zero order wavefunction of Eq.
(19.6a) by the square root of a Lorentzian of width F;

398 Rydberg atoms If weassume that �islarge enough that we may

approximatef(r2) by2/r2, then 2 Hi = 3^i(cos 012) + 4^2(cos 012).. � ,
(19.10) 7 7 and the leading term in the matrix element of Eq. (19.8) is the
dipole term. Explicitly, V e.�t�1 ^ | (19.11) The quadrupole term of Hx
cannot couple the 6pn� states to the 6s and 5d continua, and we ignore the
octopole term which couples the 6pn� state to the 5det" continua. The
matrix elements of Eq. (19.11) can bebroken into angular and radialparts.
The angular parts can be evaluated asthe scalar products of twotensor
operators using the methods of Edmonds.3 The radial parts consist of the
Ba+ ^plr^s) or (6p|r1|5d) radial matrix element and the hydrogenic (nl \
l/r^ef') radial matrix elements. Since some of the angular matrix elements are
always of order one, others being smaller, and the Ba+ radial matrix
elements donot depend on n or�, allof the dependence on the quantum
numbers nand �entersviathe radial matrix element coupling the Rydberg n�
state to the continuum. The radial matrix element of Vr\ is most sensitive
tothe smallrpart ofthe nt Rydberg wavefunction and the e�' continuum
wavefunction, and the only significant n dependence is through the l/n3/2
normalization factor of the bound Rydberg wavefunction, assuming n �
�.When the matrix elements are squared to givethe autoionization rate, we
find the autoionization rate has a 1/n3 dependence. Explicitly, ToclM3.
(19.12) While it is straightforward to show that the autoionization rates scale
as 1/n3, it is not possible to make an equally simple argument as to how they
scale with �. Consider atypicalmatrixelement ofinterest, (n(,\l/r%\e(� +
1)), whichdependson the small r part of the wavefunctions. How does the
matrix element change if n is kept fixed and � is varied from ��n to � = n
� 1? Increasing � increases the classicalinner turningpoint r{oftheouter n�
electron'swavefunction accordingto r , - ^ , (19.13) and as aresult both
thebound nt andcontinuum �(�+ 1) wavefunctions are found at increasing
r2 as � is increased. Accordingly, the matrix element of 1/rf

Experimental methods 399 decreases rapidly with increasing �. Exactly

how the matrix element decreases with �depends on the energy e, especially
as� �>n. In the example described above the autoionization of Ba 6pn�
states to the 6s and 5d states of Ba+ is by means of a dipole coupling. The Ba
5dn� states, however, autoionize by a quadrupole coupling. The quadrupole
processes are often as strong as dipole processes, counter to the intuition
developed from optical transitions. Inspecting Eq. (19.10) we can see that
the factor by which multipole processes of adjacent orders differ isrllr2.
This factor is approximately the radius of the ion divided by the inner turning
point of the Rydberg electron's orbit. For low t states this ratio is of order
one. For an optical transition the analogous factor is r/A, where X is the
wavelength of the light, and for visible radiation rlX ~ 10~3. Autoionization
can be viewed as a scattering process, and a simple classical picture leads to
the same n and � scaling as the quantum mechanical treatment given above.
The basic mechanism responsible for autoionization is superelastic scattering
of the Rydberg electron from the excited Ba+ ion to produce a ground state
Ba+ ion and a more energetic electron. Consider a single passage of the
electron past aBa+ ion in the 6pstate. For the electron to induce the transition
to the Ba+ 6s state it must come close enough that its field components are
strong enough and high enough in frequency to induce the Ba+ 6p ��
6stransition. How close the electron comes to the Ba+ and how rapidly it
goes past the Ba+ depend mostly on the �of the electron relative to the Ba+.
Increasing �decreases the field from the electron at the Ba+ ion and lowers
the frequency components of the field, thus reducing the probability of
superelastic scattering, and of autoionization. We have thus far considered
the probability of superelastic scattering on a single orbit. To obtain the
scattering rate, or autoionization rate, we simply multiply this probability by
the orbital frequency, 1/n3.4 Once again wefindthat F oc \/n3 and that T
decreases with increasing �. The scattering description we havejust given is
a twochannel description. This picture,when manychannels are present,
forms the basis of multichannel quantum defect theory.5 Experimental
methods The classic method of studying autoionizing states is vacuum
ultraviolet (vuv) absorption spectroscopy using the continuum radiation from
either a lamp, or more recently, a synchrotron in conjunction with a
spectrometer.6 This approach is straightforward in principle and generally
applicable, but it is limited to the observation of singlephoton transitions.
Usually the sample of absorbing atoms is composed of ground state atoms.
However, it is possible to use alaser to produce

400 Rydberg atoms substantial populations in excited states and observe the
absorption from them.7 An alternative approach, employed by Garton etal.
was to measure the absorption spectrum of Ba heated to high temperatures in
a shock tube.8 In addition to producing excited states the shock heating also
produced ions and electrons leading to the initial observation of forced
autoionization.8 While vuv absorption spectroscopy is relatively
straightforward, one of its inherent properties is that both the doubly excited
autoionizing state and the degenerate continuum havenonzero excitation
amplitudes from the ground state, and these amplitudes can interfere
constructively or destructively. Typically, if the interference isconstructive
on the high energy side of the autoionizing state it is destructive on the low
energy side and vice versa. The interference term between the two excitation
amplitudes changes sign across the autoionizing resonance due to the
continuum phase shift of iton goingthrough the resonance.2 The resulting
asymmetric line profiles are termed Beutler-Fano profiles, and we shall
discuss them more quantitatively in the next chapter. For the moment we note
that while the line profiles of autoionizing states are not symmetric, it isstill
possible to extract the energies and widths of the autoionizing states from
thevuv spectra. As arepresentative example, weshowin Fig. 19.2 the
spectrum of Ba obtained by vuv absorption spectroscopy.9 The light from a
microwave excited Kr or Xe continuum lamp is focused on the slits of a
spectrograph by a spherical mirror. Between the mirror and the spectrograph
the light passes through a King furnace containing Bavapor. Inessencethe
Kingfurnace consists of atubewhichisheated to 900-1100�C, over the
central 120 cmof itslength.10 In the central region theBa vapor pressure is2-
20 Torr, and 1-10 Torr offlowingAr ispresent to confine the Ba vapor to the
heated zone, keeping it away from both the mirror and the spectrograph slits.
Several features typical of spectra in the autoionizing region are apparent in
Fig. 19.2. First, there is substantial continuum absorption. Second, the
autoionizing Rydberg series converging to both the Ba+ 6p1/2 and 6p3/2
limits are clearly visible. Since the spectrum is from the ground, 6s21S0,
state the autoionizing states of Fig. 19.2 are odd parity 6pj ws1/2 and 6pjftdj
J = 1states. Third, the lineshapes are very asymmetric. Each autoionizing
state appears not as an additional absorption peak added to the continuum
absorption but as something more similar to a dispersion lineshape. In spite
of the dispersion like lineshapes, which are Beutler-Fano profiles, the
regularity of the Rydberg series converging to the 6p3/2limit is apparent, and
it is possible to determine the quantum defects and widths of the autoionizing
states converging to the 6p3/2 limit. On the other hand, below the 6p1/2
limit, where series converging to both the 6p1/2 and 6p3/2 limitsarepresent,
the spectrum is irregular duetotheinterseries interaction. Such spectra are
particularly difficult to analyze because both the bound-continuum excitation
interference and the interseries interactions lead topeculiar lineshapes, and it
is difficult to disentangle the two effects present in the spectra.

& k i i f i i B � Q O .a 1 � E � O I o s ;JI i i uojidjosqv > E

402 Rydberg atoms 6pj15d 4558A 4935A 6s15d 1 D 6s6p 6 s " So Fig. 19.3
Baenergy levelsshowingtheexcitation oftheBa6pj15d states.Thefirsttwo
lasers excite the outer electron and the third laser excites the inner electron to
the 6p!/2 or 6 state when it istuned to 4935A or 4558A. A laser method
which has been used to study autoionizing states of alkaline earth atoms is the
isolated core excitation (ICE) method first used by Cooke et al. to study the
autoionizing 5pn( states of Sr.11 It has since been used to study autoionizing
states of Mg, Ca, and Ba as well.12"14 The idea of the method is illustrated
by the study of the Ba 6pnd states. Using two pulsed lasers tuned to the Ba
6s2 ^Q-* 6s6p 1 P1 and 6s6p ^ -> 6sl5d 2 D2 transitions Ba atoms in a
thermal beam are excited to the long lived, bound 6snd l T>2 state, as shown
by Fig. 19.3. The autoionizing 6p1/215d or 6p3/2 15d states are observed by
scanning the wavelength of the third laser, at 4935 A or 4558 A, across the
6sl5d �> 6pjl5d transition while monitoring the ions or electrons resulting
from the decay of the autoionizing 6pjl5d state. An example of a typical
spectrum is shown in Fig. 19.4, the spectrum of the excitation from the Ba
6sl5d X D2 state to the Ba 6p3/215dj / = 3 state, obtained with all three
lasers circularly polarized in the same sense.15 As shown by Fig. 19.4 the
lineshape is approximately Lorentzian, as expected from Eq. (19.9). The
width may be measured directly, and the energy of the state is obtained by
adding the laser frequency at the center of the resonance to the known energy
of the 6sl5d l T>2 state. In most cases the width of the autoionizing levels is
due predominantly to autoionization. There is, however, a radiative
contribution to the width. For the Ba 6pn� Rydberg states the dominant
source of radiative width isthe radiative decay of the inner electron from 6p
to 6s, which has a width of 26 MHz.1617 As shown by Fig. 19.4 the
observed profile directly reflects the spectral profile of the autoionizing
state, asimplied by Eq. (19.9), and there isevidently negligible

Experimental methods 403 PHOTON ENERGY hw (cm"1) 21,920 21,940

21,960 21,980 63,230 63,250 63,270 63,290 TERM ENERGY (cm"1) Fig.
19.4 Low power scan of the third laser across the Ba 6sl5d 1 D2 ->
6p3/215d/ = 3 transition taken with allthree lasers circularly polarized inthe
same sense.Thisscanyields the position and width of the 6p3/215d/ = 3 state.
The narrow peak denoted by B is a coincidence with the 6s6p1P1 -> 6sl0d X
D2transition. The term energy of the 6sl5d 1 D2 state is 41315.5 cm"1 (from
ref. 15). excitation of the degenerate continuum; only the doubly excited state
is excited. In the 6sl5d�> 6pl5d transition the inner electron undergoes the
Ba+ 6s�> 6p transition, the Ba+ resonance linetransition, and the outer 15d
electron remains a spectator, making minor readjustments initsorbit to
account for the difference in the quantum defects of the 6snd and 6pnd states.
This process is shown schematically in Fig. 19.5, which shows the orbits of
the two valence electrons in the 6s2, 6sl5d, and 6pl5d states.18 In the
excitation from the Ba 6snd to the6pnd states the quantum defects of the
bound and autoionizing ndstates differ by ~0.1, inwhich case only asingle
strong peak is observed inthe spectrum.1115 However, when the quantum
defect of the autoionizing series differs from that of the bound state by 1/2,
two strong transitions from the bound n( state to the nt and (n + 1)� states
are observed.1115 Why the continuum excitation is negligible is apparent if
we consider the two alternative ways the 6sl5d atom can absorb the visible
photon required to reach the energy of the 6pl5d state. Thefirstwayis the
inner electron 6s�> 6ptransition with the outer electron remaining a
spectator. The Ba+ 6s �> 6ptransition has an oscillator strength of one17
spread over the 10 cm"1 width of the 6pl5d state, yielding df/dW ~ 104. In
general d//dWfor this excitation scales asn3, due to the widths of the
autoionizing states. The second excitation possibility is for the Rydberg
electron to be directly photoionized bythe visible photon. The Rydberg
404 Rydberg atoms 6s2 LASERS 1 a 2 6s 15d LASER 3 6p 15d Fig. 19.5
Schematicpicture ofthe Ba 6pl5dexcitationprocess. Lasers1 and 2 excite one
of the 6s electrons to the 15dstate. Laser 3 drivesthe remaining 6s electron to
the 6p state, whilethe 15d electron isa spectator (from ref. 18). electron is
most likely to be found at its outer turning point, far from the ionic core,
where the spatial variation in the wavefunction is slow. In contrast, at the
energy of the 6pnd state the continuum wavefunction exhibits rapid spatial
variation far from the ionic core, and the large r parts of the wavefunctions
contribute littletothephotoionization matrixelement. Onlythesmall rpart ofthe
Rydberg wavefunction contributes to the matrix element. Accordingly,
d//dWfor direct photoionization of the 6snd states scales as1/n3, duetothe
normalizationof the Rydberg wavefunction near the core. Thus the ratio of the
two oscillator strengths per unit energy for the inner and outer electron
absorptions scales as n6. For the direct photoionization of the 15d state
df/dW ~ 10~4 so theratioofthetwo oscillator strengths is ~108. Even the
interference cross term is a factor of 104 smaller than the peak cross section.
While the isolated core excitation method makes the analysis of single
isolated autoionizing states trivial, its real strength lies in unravelling the
spectra of interacting series. One example of the interaction between
different series of autoionizing states is the interaction among the Ba / = 1
odd parity states converging tothe6p1/2and 6p3/2limits.In thevuv spectrum
ofFig. 19.2, theregion below the 6p1/2 limit exhibits strong interseries
interaction, but due to the added complication of attendant continuum
excitation it is difficult to separate the effect of interseries interaction from
the bound-continuum interference in the excitation. As we shall show in the
following chapter, using the ICE method it is straightforward to examine the
effects of interseries interaction with negligible

Experimental methods 405 DISK COMPUTER N2-LASER DYELASER IE

DYELASER DYELASER I EXCIMER LASER Fig. 19.6 A schematic view
of an apparatus for measuring photoexcitation cross sections
andphotoelectron energyand angulardistributions.Theatombeamcomes
outofthepage, and Dj and D2 are the electron and ion detector, respectively
(from ref. 25). interferences in the excitation amplitudes.419 It has also
proven useful for experiments involving multiphoton excitation of the inner
electron.20"24 Other strengths of the ICE method derive from the fact that the
atoms are in a low density beam, not a high pressure absorption cell. As
shown in Fig. 19.1, an atom in a Ba 6prc�state can autoionize into the
degenerate continua above lower lying states of Ba+. By analyzing the energy
and angular distributions of the ejected electrons itispossibletodetermine
thebranching ratiosfor autoionization tothepossiblefinalstatesofBa+
andtheejected electron. Suchmeasurements are far more stringent tests of
calculations than measurements of total autoionization rates. The apparatus
used to make the electron spectroscopy measurements is not much more
complicated than that required to make the total photoexcitation cross section
measurements. In Fig. 19.6 we show the apparatus used by Sandner et al.25
to measure the energy and angular distributions of electrons ejected from the
Ba 6pns states. The electron spectrometer of Fig. 19.6 is a time of flight
spectrometer, which has a 50meV resolution at electron energies of 1 eV. We
have described the excitation from the bound Ba 6snd to autoionizing 6pnd
states, and it ispossible by purely optical means to excite the 6sn� states of
�< 4.

406 Rydberg atoms I5 CD l5 Reid Fig. 19.7 Schematic diagram of the Stark
switching technique applied to n = 7 Rydberg states. The arrow shows laser
excitation of the Stark state in thefieldEo. Thefieldis then reduced to zero
adiabatically to produce the �= 5 state. The zerofieldseparations of the �
states are exaggerated for clarity. It ispossible to take advantage of the long
lifetimes and large dipole momentsof the bound 6sn� Rydberg states to
excite bound 6sn� Rydberg states of arbitrary� and observe the transitions
to their autoionizing 6pn� analogues. To produce higher �states a Stark
switching technique, originally proposed by Freeman and Kleppner,26 can
be used. It was first used by Cooke etal.n to populate Sr 5sn( states of �up
to 7. The basic idea is shown in Fig. 19.7, a schematic energy level diagram
of the n � 7, m � 0 Stark manifold as afunction of staticelectricfield.For
claritythe zerofieldseparations of the � states are exaggerated. At thefieldEo
the Stark levels are well resolved and a single Stark state can be excited
with a laser. Thefieldisthen reduced to zero adiabatically, and the initially
excited Stark level evolves into a single zerofieldangular momentum state. In
Fig. 19.7 weshow the laser excitation of the Stark state which evolves
adiabatically into the zero field ( = 5state as thefieldisturned off. The �state
produced is determined bywhich Stark state is initially excited with the laser.
Thefirstapplication of the Stark switchingtechnique to a wide range of
�states was made byJones and Gallagher who produced the Ba6sn(statesof
11 < n < 13 and � > 4.16 They initially excited the Ba atoms from the
ground 6s21S0 state to the 6s6p1P1state and then to a 6snkStark state in
afieldof 3 kV/cm. Thisfieldis strong enough that the Stark levels are ~2 cm"1
apart and there issufficient6snd character in each 6snk Stark level for
excitation. Thefieldwas reduced to zero in 2-3 jus. These times are
determined by two limits. The zero field separation between the highest t
states is 200 MHz, setting a lower limit on the switching time. On the other
hand, if the fieldis switched too slowly, black body radiation

Experimental methods 407 drives population to other states, the atoms in �~

4states decay radiatively, and the atoms pass out of the interaction region.16
This method has also been used by Pruvost etal.21tostudy Ba 6pn� and
6dn�states of high�and by Eichmann et al28 to study Sr n'gn� states of
high �. An alternative technique, originally proposed by Delande and
Gay,29 and first used byHare etal30 hasbeen used byRoussel etal31
toproduce atomsin circular Ba (� = m = n� 1) 6sra� states for subsequent
excitation to autoionizing 6pn� states. The atoms are excited in a strong
electricfieldto the highest energy 6snk Stark state and pass adiabatically into
a region where there is only a modest magnetic field perpendicular to the
direction of the electric field. The highest energy Stark state is one which is
originally a Stark state of m = 0 in the electric field but which evolves into
the m = n � 1 state in the magnetic field. Measurement of the autoionization
rates by sweeping the laser wavelength across the bound �� autoionizing
state transition isamethod whichworkswell for autoionizing states of widths
large compared to the laser linewidth. For high � autoionizing states this
condition is not satisfied, and it is necessary to use the depletion broadening
approach first used by Cooke et al32 The method takes advantage of two
facts. First, the laser lineusuallyhasan approximately Gaussian profile, which
has line wingswhich drop more rapidly than those of a Lorentzian, whilethe
autoionizing statehasaprofile whichis aLorentzian. Thus,at adetuning from
line center much larger than the laser linewidth the observed signal is
determined by the Lorenztian tail of the cross section at the detuned laser
wavelength, not by the light in the wing of the laser line at the maximum of
the crosssection. In thiscase,when the laser is tuned far from linecenter,
thewidth of the laser can be ignored. The second important fact is that since
the optical cross sections are so large, it is possible to observe strong signals
even when the detuning is such that the cross section is two or three orders of
magnitude below the peak cross section. For the optical transition from the
bound 6sn( state to the autoionizing 6pn� state the optical cross section is
given by the Lorentzian form nr (19.14) where Aco is the detuning from
linecenter, Fn�is theFWHMdueto autoionization and radiative decay and fl
is proportional totheionic 6s-6p dipolematrix element squared. Now
consider the 6sn'�'�> 6pn'�' transition for �'�^ ( or n' ^ n. The cross
section on>z< (Ao>) is obtained from on�(Aa>) by replacing Tn� by
Tn>�>. If the sample of No bound 6sn� Rydberg atoms is exposed to an
integrated photonflux4>, then the number of atomsexcited tothe autoionizing
statesisgiven by Nne = Ntion& (19.15) if onfo � 1. However, with available
laser powers the regime on( <E>�1 iseasily reached, in which case

408 Rydberg atoms Nn� =JVO(1 - e~^(Aw)o). (19.16) If the flux of the
laser is high enough that cr^(O) <& � 1, the observed signal is broadened to
awidthsubstantially in excess ofthewidthof thecross section of Eq. (19.14).
We assume that we can measure directly the width of the 6sn� �> 6pn�
transition, i.e. that F�exceedsthe laser linewidth. Wealsoassume that we
canuse a high enough laser power so that both the6sn� �> 6pn� andthe
6sn'�'�> 6pn'�' transitions are broadened to awidth substantially in excess
ofthe laser linewidth. In this case, if we compare thedetunings A&vand
A(wn^ atwhich thesignalsare half their maximum values, from Eq. (19.16)
we canseethat ont(Aa)nt)<!> = on'�>(Aa>n'�>)<& = In 2. (19.17)
Dividing by the common laser flux of <E> yields on�\DL(ji)n�) = on>�>
(Aa)n>�>). (19.18) Using the form of the cross sections given by Eq.
(19.14), Eq. (19.18) maybe written as (Aav)2 + (Tne/2)2 Since the detuning
must be large compared to the laser linewidth, if we are measuring awidth
Yn.p less than thelaser linewidth, Eq. (19.9) reduces to r (Aa)n�y+ (rn�ny
It isworth notingthat there isnothing special about thepointNn�/N0 = 111,
where the signal is equal to half the peak signal. We could equally well have
chosen Nn�/N0 = 1/3 or any other value, but at these values of Nn�/N0 it
is harder to determine Ao> experimentally. Experimental observations of
autoionization rates Earlier in this chapter we presented arguments to show
that the autoionization rates of, for example, the Ba6pn� states should scale
as1/n3, provided that n � �. Thispoint isshown rather clearly in Fig. 19.8,a
plot ofthe observed widths of the three Ba 6pj nd,/ = 3series.33Thetwo
6p3/2ndJ = 3 series werelabelled + and by Mullins etal.34since they are
reached from the bound 6snd *D2 and3D2 states, respectively, using isolated
core excitation. The 1/n3 dependence is quite evident for all three series,
asisthedeviation of the 6p1/2nd5/2widthsatn* =8and 17. At these values of
n* the interaction with approximately degenerate 6p3/2ftd states altersthe
1/n3 dependence of the widths.There are,infact, manyexamples of the IIn3
scalingof the autoionization rates,but itis important to keep inmind that this
scaling law is only valid for n � �. Whenn ~ �the autoionization rates of
an �

Experimental observations of autoionization rates 409 100 n 1 1�i�i�i r

16 18 20 Fig. 19.8 Autoionization rates of the three 6pnd J = 3 series
6p1/2nd5/2( A), 6r^/2nd+ (�), and 6p3/2nd~(M). The 6p3/2rcd+ and
6p3/2rcd~ states are excited from the 6snd X D2 and 6snd 3 D2 states
respectively. The perturbations in the 6p1/2nd5/2 series are quite evident
(from ref. 33). seriescanincreasewithn. This phenomenon is observed
intheBa4fngseries.The 4f6g state is broader than the 4f5g state.35 Part of the
reason is that for n = � + 1 the nt wavefunction hasonly one lobe inthe
minimum of the combined coulombcentrifugal potential. For n = � + 2 the
inner turning point moves to noticeably smaller radius. For example, the
inner turning point of the 5g state is at 13.Sa0, while it is at 12.0a0 for the 6g
state. As n is further increased, the inner turning point continues to move to
smaller radius, but very slowly, and the effect of the l/n3/2 normalization
factor of the ng wavefunction quickly becomes more important. Another
possible contributing factor is wavefunction collapse. In La the 4f
wavefunction is collapsed into an inner well in the radial potential, while in
Ba, with asmaller nuclear charge, the 4felectron isjust onthevergeof
collapsing.36'37 It maybethat thepresence of the 6g electron at asmaller
orbital radiusscreensthe 4f electron from the nuclear charge more efficiently
than does the 5g electron. Thus collapse of the 4f electron is more likely in
the 4f5g state than in the 4f6g state. The partial collapse of the 4f
wavefunction in the 4f5g state might lead to a smaller autoionization rate for
the 4f5g state than for the 4f6g state. Connerade has pointed out that it may be
possible to control externally the extent to which a wavefunction
collapses.38 These observations may support his suggestion, but careful
calculations need to be carried out to verify this point. Autoionization rates
also decrease rapidly with �, a point first shown experimentally by Cooke
et al.n who measured the autoionization rates of the Sr 5pl5� states of � =
2-7. The �> 2 states were populated using the Stark switching

410 Rydberg atoms 10 E 10 cd (L) cd K >�> cd o Q o CD 1 10 -1 10 "2 10

"3 10 "4 10 "5 - o : 0 O : In n = r 2 o \ 2 A \ A i ) A 9 10
eFig.19.9Plotofscaledtotaldecayratesn3TofBa6p 1/2n� J = �4- 1
autoionizing states in atomic units vs �. For � = 0-4 the measured rates (O)
shown are the average rates from many n values. The data for the rates for
�> 4 are for n = 12. The solid line is a simple theoretical calculation based
on the dipole scattering of a hydrogenic Rydberg electron from the 6pcore
electron. Note that the core penetration of the lower �states reduces the
actual rate from the one calculated using the dipole scattering model. The
constant total decay rate for �>8 is the spontaneous decay rate of the Ba+ 6p
state (from ref. 39). technique described earlier. Their measurements showa
dramatic decrease of the observed autoionization width from 15 cm"1 at � =
2 to 1 cm"1 at t = 5. The measurements for �>5 were limited by the
linewidth of the laser. The widthsof the Ba 6p1/2n� states of11< n < 13 and
4 < �< n - 1 were measured by Jones and Gallagher16 using both the Stark
switching technique26 and the depletion broadening technique to measure
decay rates below the laser linewidth.32 Their results showed clearly the
rapid decrease of the autoionization rate with �, out to �~9 where the
autoionization rate falls belowthe radiative decay rate of the Ba+ 6p1/2 ion,
and the total decay rate is dominated bythe radiative decay rate, which is
independent of n and �. In Ba, the autoionization rates of many 6pn( states
have been measured,41619'27'34'39^2 and it is interesting to combine the
autoionization rates of the high �states,16 with those measured for low
�states. In Fig. 19.9we show the measured scaled autoionization rates, n3T,
for the 6p1/2n� states of/ = �+l.39 The measured values are compared to
the values calculated using Ba+ coulomb wavefunctions and hydrogenic
wavefunctions, i.e. the approach outlined earlier. The calculations show
uniformly decreasing autoionization rates with increasing
Electron spectroscopy 411 10 Continuum Channels 6sns^ \ 6S6P 1 P � - 5
Fig. 19.10 Energy level diagram for the autoionization of Ba 6p5/2ns1/2 / =
1 states. As shown, above the 6p1/2limit there are ten continuum channels.
Below the 6p1/2 limit there are only eight continuum channels. The two step
laser excitation of the spherically symmetric 6sns1S0state isalsoshown. Not
shownis the laser excitation from the 6sns state to the 6p3/2�s1/2 state
(from ref. 25). �.For � > 3the calculated and observed rates agree. For�<
3they do not. The disagreement is not surprising since the �< 3 states have
significant core penetration, and the calculation is almost certain to be wrong
for �< 3. A more surprising observation is that the measured 6p1/2nf
autoionization rate is the highest, substantially higher than the autoionization
rates of the t < 3 states, which we might intuitively expect to be larger. A
possible reason why this is sois that the �<3 states are all distinctly core
penetrating, while the ni states are not. In other words, the inner turning point
of the ni wavefunction approximately coincides with the outer turning point
of the inner 6p electron's wavefunction. In this case the two wavefunctions
overlap where both electrons are moving slowly, and therefore interacting
strongly. Electron spectroscopy If wereplace the schematic energy level
diagram of Fig. 19.1 byFig. 19.10, which more accurately represents the Ba
atom, we can see that the Ba 6p3/2�s1/2 / = 1

412 Rydberg atoms Fig. 19.11 (a) Time of flight spectra of Ba 6p3/2�s J = 1
autoionization electrons. Peak 1, decay into the Ba+ 6s continuum; peak 2,
into the Ba+ 5d3/2)5/2 continua; peak 3, into the Ba+ 6p1/2 continuum. The
drift length was approximately 10 cm. The spectra were recorded with a gate
width of 14 ns. (b) The same as (a) but with a drift length of approximately
45 cm and a gate width of 6 ns (from ref. 25). atoms,for example, can
autoionize intoten continua of different �abovethe 6s1/2, 5dj, and
6p1/2ionization limits.25Below the 6p1/2limit the Ba 6pjns1/2 J = 1 states
can only ionize into the eight continua above the 6s1/2 and 5dj ionization
limits. The allowed final states have been probed by measuring the energy
and angular distributions of electrons ejected from the Ba 6pj ns1/2 J = 1
states. There are two quantities of interest for each autoionizing state, the
branching ratios to the final ion states, and the anisotropy parameters of the
angular distributions of the ejected electrons inthe autoionization to specific
ionic states.When the autoionizing Ba 6pj ns1/2 / = 1states are excited from
the spherically symmetric 6sns 1 S0 states by electric dipole excitation, the
angular distribution of the electrons ejected at any energy takes the form43
7(0) = /o[l+i8P2(cose)] (19.21) where 6 is the anglebetween the Evector of
the exciting laser and the momentum vector of the escaping electron and /? is
the anisotropy parameter. An interesting point about the ICE method isthat
since there isno excitation of the continuum, which has a varying phase, the fi
parameter and the branching ratio are constant across the autoionizing state's
profile. Conservation of momentum requires that virtually all the kinetic
energy of autoionization appears in the electron. Thus thefinalstate of the ion
is implied by the energy of the ejected electron, and measuring the energy and
angular distributions of the ejected electronsyields thebranchingratiosfor
autoionization into the possiblefinalstates of the ion. In Fig. 19.11 we show
the time of flight spectra, obtained with 0 = 0, for electrons from a 6p3/2
ns1/2 state lying above the 6p1/2 ionization limit.25 With a 10 cmflightpath
for the electrons, the electrons ejected in autoionization to the Ba+ 6s1/2 and
5dj states are barely resolved, while with the higher resolution afforded by a
45 cmflightpath the electrons from autoionization to the 6s1/2 and 5dj states
arewellresolved, showinglittle autoionization totheBa+ 5d5/2 state. To

References 413 convert the time offlightspectra, such as those shown in Fig.
19.11, to branching ratios, they must be either measured at all values of the
angle O and integrated over 0, or measured at the magic angle, G = 54.7�,
where P2(cos �)= 0. Rotating the polarization of the laser driving the 6sns 1
S0-^ 6pyns1/2 J=l transition yields the angular distributions of the ejected
electrons, and they have been measured bytwogroups.25'44'45 Angular
distributions from the autoionizing states of Sr, Ca, and Mg have also been
reported.46"48 An interesting point to note in Fig. 19.11(b) is that a
significant number of the observed electrons are from autoionization to the
Ba+ 6p1/2 state. In fact, 40% of the Ba 6p3/2fts1/2 atoms autoionize to the
6p1/2state of Ba+, morethan to any other single ion state,25 in spite of the
fact that it is a quadrupole process, whereas autoionization to either the Ba+
6s1/2 or 5djstates occurs by adipole process. The preferential population of
the Ba+ 6p1/2state when Ba 6p3/2np states of n > 12 are populated by a
laser has been used by Bokor et al. to make a laser on the Ba+
6P1/2��5d3/2transition at 600nm and the Ba+ 6p1/2�> 6s1/2transition
at493 nm.49 References 1. W. Sandner, Comm. At. Mol. Phys.20, 171
(1987). 2. U. Fano, Phys.Rev. 124,1866 (1961). 3. A. R. Edmonds, Angular
Momentum in Quantum Mechanics, (Princeton University Press, Princeton,
1960). 4. W. E. Cooke and C. L. Cromer, Phys.Rev. A 32,2725 (1985). 5. M.
J. Seaton, Rep. Prog. Phys.46, 167 (1983). 6. J. Berkowitz, Photoabsorption,
Photoionization, and Photoelectron Spectroscopy (Academic Press, New
York, 1979). 7. J. L. Carlsten, T. J. Mcllrath, and W. H. Parkinson, /. Phys.B
8, 38 (1962). 8. W. R. S. Garton, W. H. Parkinson, and E. M. Reeves, Proc.
Phys.Soc. London 80, 860 (1962). 9. C. M. Brown and M. L. Ginter, /. Opt.
Soc. Am. 68,817 (1978). 10. C. M. Brown, R. H. Naber, S. G. Tilford, and
M. L. Ginter, Appl. Opt. 12,1858(1973). 11. W. E. Cooke, T. F. Gallagher, S.
A. Edelstein, and R. M. Hill, Phys. Rev. Lett.40, 178 (1978). 12. G. W.
Schinn, C. J. Dai, and T. F. Gallagher, Phys.Rev.A 43, 2316 (1991). 13. V.
Lange, V. Eichmann, and W. Sandner, J. Phys. B 22, L245(1989). 14. L. D.
von Woerkem and W. E. Cooke, Phys.Rev. Lett. 57,1711 (1986). 15. N. H.
Tran, P. Pillet, R. Kachru, and T. F. Gallagher, Phys.Rev. A 29,2640 (1984).
16. R. R. Jones and T. F. Gallagher, Phys. Rev. A 38,2946 (1988). 17. A.
Lindgard and S. E. Nielson, At. Data Nucl. Data Tables 19,613 (1977). 18.
T. F. Gallagher, /. Opt. Soc. Am. B 4794(1987). 19. F. Gounand, T. F.
Gallagher, W. Sandner, K. A. Safinya, and R. Kachru, Phys.Rev.A 27, 1925
(1983). 20. L. A. Bloomfield, R. R. Freeman, W. E. Cooke, and J. Bokor,
Phys.Rev. Lett. 53,2234 (1984). 21. P. Camus, P. Pillet, and J. Boulmer, /.
Phys. B 18, L481 (1987). 22. N. Morita, T. Suzuki, and K. Sato, Phys.Rev. A
38, 551 (1988). 23. U. Eichmann, V. Lange, and W. Sandner, Phys. Rev. Lett.
64, 274 (1990). 24. R. R. Jones and T. F. Gallagher, Phys. Rev. A 42, 2655
(1990). 25. W. Sandner, U. Eichman, V. Lange, and M. Volkel, /. Phys.B 19,
51 (1986).

414 Rydberg atoms 26. R. R. Freeman and D. Kleppner, Phys.Rev. A

14,1614 (1976). 27. L. Pruvost, P. Camus, J.-M. Lecompte, C. R. Mahon, and
P. Pillet, J. Phys.B 24,4723 (1991). 28. U. Eichmann, V. Lange, and W.
Sandner, Phys.Rev. Lett. 68,21 (1992). 29. D. Delande and J. C. Gay,
Europhys. Lett. 5, 303 (1988). 30. J. Hare, M. Gross, and P. Goy, Phys. Rev.
Lett. 61, 1938 (1988). 31. F. Roussel, M. Cheret, L. Chen, T. Bolzinger, G.
Spiess, J. Hare, and M. Gross, Phys.Rev. Lett. 65,3112(1990). 32. W. E.
Cooke, S. A. Bhatti, and C. L. Cromer, Opt. Lett.7, 69 (1982). 33. T.F.
Gallagher, in Electronic andAtomic Collisions, eds D.C. Lorents, W. E.
Meyerhof, and J. R. Peterson, (Elsevier, Amsterdam, 1986). 34. O. C.
Mullins, Y. Zhu, E. Y. Xu, and T. F. Gallagher, Phys.Rev. A 32,2234 (1985).
35. R. R. Jones, P. Fu, and T. F. Gallagher, Phys.Rev. A 44, 4260 (1991). 36.
M. G. Mayer, Phys.Rev. 60, 184 (1941). 37. D. C. Griffin, K. L. Andrew, and
R. D. Cowan, Phys.Rev. Ill, 62 (1969). 38. J. P. Connerade, /. Phys.B 11,
L381 (1978). 39. R. R. Jones, C. J. Dai, and T. F. Gallagher, Phys.Rev. A 41,
316 (1990). 40. J. G. Story and W. E. Cooke, Phys. Rev. A 39, 5127 (1989).
41. X. Wang, J. G. Story, and W. E. Cooke, Phys.Rev. A 43, 3535(1991). 42.
B. Carre, P.d'Oliveira, P. R. Fournier, F. Gounand, and M. Aymar, Phys.Rev.
A 42(6545) 1990. 43. C. N. Yang, Phys.Rev. 74, 764 (1948). 44. W. Sandner,
R. Kachru, K. A. Safinya, F. Gounand, W. E. Cooke, and T. F. Gallagher,
Phys. Rev. A 27, 1717 (1983). 45. R. Kachru, N. H. Tran, P. Pillet, and T. F.
Gallagher, Phys.Rev.A 31, 218 (1985). 46. Y. Zhu, E. Y. Xu, and T. F.
Gallagher, Phys.Rev. A 36, 3751 (1987). 47. V. Lange, U. Eichmann, and W.
Sandner, /. Phys.B 22, L245(1989). 48. M. D. Lindsay, L.-T. Cai, G. W.
Schinn, C.-J. Dai, and T. F. Gallagher, Phys.Rev.A 45, 231 (1992). 49. J.
Bokor, R. R. Freeman, and W. E. Cooke, Phys.Rev. Lett. 48,1242(1982).

20 Quantum defect theory In the previous chapter we considered isolated

autoionizing states coupled to degenerate continua. The perturbative
treatment presented there is perfectly adequate to describe isolated
autoionizing states but is at best awkward for the description of anentire
seriesofinteracting autoionizing series.A straightforward way of describing
these phenomena is to use quantum defect theory (QDT), a multichannel
scattering approach first developed by Seaton.1 Depending on the choice of
basis functions the QDT equations can take many forms.2 Two choices are
commonly used. The first is based on the separated ion and electron and uses
reactance, or R, matrices.1"7 The second is based on the normal modes of
the short range electron-ion scattering.8"10 In this chapter we briefly
describe the essential ideas of QDT, following the approach of Cooke and
Cromer.3 This approach has the advantage of showing the connection
between the two formulations. The essential notions presented here are
adequate to understand many of the subtle features of the spectra of
autoionizing states, but for a more complete description the reader is referred
to the original articles.1"10 Quantum defect theory (QDT) In the second
chapter we described a Rydberg electron outside a spherically symmetric
ground state ion core using single channel QDT. Only asingle channel is
necessary since the ionic core cannot exchange energy or angular momentum
with the Rydberg electron. An implicit assumption of single channel QDT
isthat allexcited statesof the ion are so far removed inenergy that
theymaybeignored. In a Rydberg state of any atom but H, when the distance, r,
of the Rydberg electron from the ion core is greater than a core radius, rc, the
potential is a coulomb potential, but for r <rc the potential is usually deeper
than a coulomb potential. The effect of the deeper potential is that if an
incoming coulomb wave scatters from the ion core, the reflected wave has
aphase shift of TIJU compared to what it would have if it scattered from a
proton. In other words the standing wavefunction for all r > rc is given by y\r
= - [f(w, �, r)cos Jtju + g(W, �, r) sinTVJU] (20.1) r 415

416 Rydberg atoms 1 2 3 Fig. 20.1 A three channel, three limit problem.
Three levels of the ion, shown by the bold lines, have Rydberg series
converging to them from below and continua above them. The interactions
between the series are shown by the horizontal arrows. Bound-continuum
interactions are shown by single headed arrows while bound interactions are
shown by double headed arrows. where/and gare the regular and irregular
coulombwavefunctions. For scattering from a proton the g wave would be
absent. The utility of QDT is that ju is nearly energy independent, a property
which derives from the fact that if rc is small, ~ la0, the kineticenergy of the
electron whenitcollideswiththeion core at r<rc is large, �10 eV and over an
energy range of <1 eV thephase shift JIJU is constant, as shown in Chapter 2.
An alternative way of expressing the wavefunction of Eq. (20.1) is � = ^ ^ [/
(W, �, r) + tan fan) g (W, t, r)], (20.2) r which resembles a scattering
wavefunction. If the ion does not have only a single spherically symmetric
state it isno longer true that the energy and angular momentum of the Rydberg
electron and the ion are separately conserved, and multichannel QDT must be
used to account for the fact that several combinations of ion and electron
energy, angular momentum, and parity may conserve the total atomic energy,
angular momentum and parity, givenby W, /, and n, respectively. Consider the
case shown in Fig. 20.1, inwhich there are three ionic states of different
energies. In Fig. 20.1we show the bound states converging to these ionic
limits and the continua above them. Each set of bound Rydberg states and
associated continuum, both having the same angular

Quantum defect theory (QDT) 417 momentum and spin, constitutes a channel,
and Fig. 20.1 depicts three such channels. In an isolated atom for the
channels to be interacting they must have the same total angular momentum
and parity. It iseasy to describe the channels when the Rydberg electron isfar
from the ionic core, i.e. as r-> oo. For example, asr�> oo? the Ba/ = 1 odd
parity channels arewelldescribed asproducts of a specific Ba+ ionic state
and an electron in a well defined state. Examples are 6s1/2�p1/2,
6si/2�p3/2, 5d3/2�p1/2, 5d3/2�p3/2, 6py2ns1/2 � � �. In each case
the relative orientation of the ionic, electron, and spin angular momenta is
specified by the total angular momentum /. These channels are usually called
the dissociation or collision channels, and we shall describe them using the
wavefunctions (p{. In the collision channels we can unambiguously partition
the total energy into an ionic and a Rydberg electron part according to W=Wi
+ Ii9 (20.3) where Wt is the Rydberg electron's energy and It is the ionic
core's energy. A channel is open if W( > 0 and closed if Wl < 0. In these two
cases the Rydberg electron is in a continuum and a bound state respectively.
Far from the ionic core the logical wavefunctions with which to describe the
collision channels are the (j)t wavefunctions given by 0. = _ [Xif(Wt �, r)
cosTtVi + XigiWi *,r) sin n vj, (20.4) r where %i is the product of the total
ionic wavefunction and the angular part of the Rydberg electron
wavefunction, including spin. In Eq. (20.4) it isapparent that vt specifies the
fraction of the regular and irregular coulomb functions. Equivalently, nvt is
the phase shift from the regular hydrogenic/function, as in Eq. (20.1).
Although the properties of the/and g functions are outlined in chapter 2, it is
worth summarizing their properties here.8 The / and g coulomb functions are
termed regular and irregular since as r�> 0,/oc re+1 andg oc r~ � . Due to
the r = 0 behavior of the g function, in H only the/wave exists. As r �� oo
for Wt > 0 the/ and g waves are sine and cosine functions, and if Wt > 0, nvx
simply specifies the phase of the wavefunction relative to the
hydrogenic/wave. If Wt < 0the/and g waves both have exponentially
increasing and decreasing parts, and, as we have seen in Chapter 2, only if
Wt = ^ \ (20.5) does the wavefunction of Eq. (20.4) vanish as r�> oo.
Unless Eq. (20.5) is satisfied the wavefunction diverges exponentially as
r�> oo if w{ < 0. Although as r �� oothe wavefunctions of different
collision channels are very different, at small rthey are similar, and it
ispossible to find normal modes of the scattering from the ionic core. Over
an energy range AW we can find a radius R, R > rc, such that for r < R the
wavefunctions of Eq. (20.4) are energy
418 Rydberg atoms independent, aside from the normalization factors of the
closed channels. If we disregard the r = ooboundary condition for the closed
channels, vxcan take any value, not just v{ � 1/V�2W;. If we in addition
apply the continuum normalization per unit energy to these wavefunctions,
then in the region rc < r < R we can think of all the 0, wavefunctions as
energy independent scattering, or continuum, wavefunctions. Our goal is to
find the normal modes of the scattering from the ionic core. The
wavefunctions of the normal scattering modes are the standing waves
produced by a linear combination of incoming coulomb wavefunctions which
is reflected from the ionic core with only aphase shift. The composition of
the linear combination is not altered by scattering from the ionic core. These
normal modes are usually called the a channels, and have wavefunctions in
the region rc<r<R given by 1 """ \ *, r) cos niia - Y UiaXig(Wi9 �, r) sin
jr/J, (20.6) where Jtjua is the phase shift analogous to ju in Eq. (20.1) and
Uia is a unitary transformation. Often juais termed an eigen quantum defect. If
we compare Eq. (20.1) to Eq. (20.6) we can see that in the former case a
single incident incoming hydrogenic wave results in a phase shifted outgoing
wave, whereas in the latter case a linear combination of incoming hydrogenic
waves results in the same linear combination of outgoing waves, only phase
shifted. Since the 0, wavefunctions are energy independent for r < R, the jua
quantum defects are also. More generally, if the wavefunctions are
slowlyvarying with energy, the values ofjuaare also. The normal modes with
the wavefunctions Va defined by Eq. (20.6) are in essence the eigenfunctions
which match the boundary condition at rc, but they do not match the r �> oo
boundary condition if there are any closed channels. In contrast the M^-
wavefunctions match the r -� ooboundary condition but not the r = rc
boundary condition. In general for r > rcthe wavefunction ^ can be written as
a linear combination of either the ^ or Wa wavefunctions. Explicitly ] a^a -
(20.7) Writing 0, and ^a explicitly using Eqs. (20.4) and (20.6) and equating
the coefficients of %/(Wi?�,r) and %ig(Wi^s) yields Ai cos jtVi = \ UiaBa
cos Jtjua (20.8a) a At sinnvt = - Y UiaBa sinjrjua (20.8b)

Quantum defect theory (QDT) 419 Multiplying Eqs. (20.8a) and (20.8b)
bysinnvt and cos Jivt respectively and adding and subtracting them yields A
= � Uia cos n(vt + [ia)Ba, (20.9a) Ba. (20.9b) Using the fact that Uia is a
unitary matrix for which UT = U"1, Eqs. (20.8a) and (20.8b) can be written
as \ UiaAt cos nvt = Ba cos Jtjua, (20.10a) V UiaAisin jiVi = -Ba sin 7tjua.
(20.10b) i Multiplying Eqs. (20.10a) and (20.10b) by cos Jijua and -sin 7tjua
respectively and adding and subtracting yields / i Uia cos 7i{vt + fiJAi = Ba,
(20.11a) Uia sin n(Vi + Ha)At = 0. (20.1lb) Eqs. (20.10a) and (20.10b)
appear in the Fano formulation of QDT.8 Following Cooke and Cromer,3 we
shall use Eqs. (20.11a) and (20.11b). Irrespective of which set is more
convenient, nontrivial solutions can only be found if det |U^ sin Jt{vt + jua)\
= 0. (20.12) Eq. (20.12) defines the quantum defect surface. For bound
states, all channels are closed, Wt < 0 for all /, and the dimensionality of the
surface defined by Eq. (20.12) is one less than the number of channels. For a
two channel problem Eq. (20.12) defines a line while for a three channel
problem it defines a two dimensional surface. From Eq. (20.12) it is
apparent that replacing vt by vx + n, where n is an integer, either leaves the
matrix unchanged or reverses the sign of an entire row. In either case Eq.
(20.12) is satisfied for the same value of vv The same reasoning may be
applied to v2, so that the quantum defect surface defined by Eq. (20.12)
repeats modulo 1in vt. Let us for a moment consider the three channel
problem depicted in Fig. 20.1. We have just discussed the case in which Wt
< 0 for all /, i.e. the region below the lowest limit. The quantum defect
surface defined by Eq. (20.12) is a two dimensional surface inscribed in a
cube of length Avt- = 1 on a side. Now let us consider the region between
thefirst and second ionization limits,where channel 1 is open. Since 0X is a
continuum wave the r �> �� boundary condition does not

420 Rydberg atoms wB CD c LU Fig.20.2 Radialpotentials due to three

ionicstates atdifferent energies.At energy WA, for r> rA only channel 1
isclassically allowed and only (pl existsinthis region. At energy WB, for r >
rB channels 1 and 2 are classically allowed and the wavefunction is an
energy dependent linear combination of 0j and 02constrain the possible
values of nvx. However, due to the interchannel couplings at smallr, only
thevalues of JTV1 specified byEq. (20.12) satisfy both the r =rc and r��
oo boundary conditions. In this case the quantum defect surface is identical to
the surface below the first limit. Now however JZV1 is only the continuum
phase shift, sometimes labeled JIT. It isalso useful to think of the problem in
scattering terms. In Fig. 20.2 wehave drawn the radial potentials for channels
1,2, and 3. At an energy WA, above the first limit, but below the second limit
for r > rA the wavefunction is composed entirely of the channel 1
wavefunction 01? given by Eq. (20.1). We can imagine putting aspherical
box of radius rA around the ioniccore and asking, "What is the normal
scattering mode for the scattering of an electron from the contents of the
box?" Sincethere is only one continuum wave, the onlyissueis its phase shift
nvx, and depending upon the proximity of the energy to the energies of the
bound states of channels 2and 3 the phase shift Jtv1, corresponding to the
normal mode, has different values. If the energy is raised above the second
limit there are two open channels. In Fig. 20.2 at an energy WB for r > rB the
wavefunction is composed of a linear combination of <j>1 and 02- Kweput
aradial boxof radiusrB around the ioniccore we can again ask, "What are the
normal modes for electron scattering from the contents of the box?" In other
words, what linear combinations of incoming coulomb wavefunctions will
suffer at most aphase shift when scattering from the contents of the box?
There are two wavefunctions, labelled by p = 1,2. They are linear
combinations of (j)1 and 02> given by iPXif( Wit �, r)cosmp + �
AipXig(Wh �, r)sin mp. (20.13) �=1,2 i=l,2

Geometrical interpretation of the quantum defectsurface 421 In Eq. (20.13)

the eigen phase shift Jtrpplays the same role//a playsin Eq. (20.6). However,
the values of rp are not energy independent due to the bound statesof channel
3. The phase shifts rp depend on the proximity of the energy to the
energiesofthe statesof channel3. Similarly,the composition ofthenormalmodes
of scattering from the box of radius rB is not energy independent, and the Aip
values depend strongly on the energy. Tofindthe values of the eigen phase
shift rp we simply replace 7ivt by nx for all the open channels in the
determinant of Eq. (20.12) and solve Eq. (20.12) for the two possible values
of T, which are rp, p = 1,2. If there are Popen channels there are Pvalues of
rp. While it isnot transparent from the discussion up to this point that this
procedure is reasonable, as we shall see, these values of xp lead to
scattering and reactance matrices which are diagonal, and continuum
wavefunctions which are the normal scattering modes. When there are
multiple values of T, the dimensionality of the quantum defect surface is
reduced. In the region below the second limit the quantum defect surface
defined by Eq. (20.12) is a two dimensional surface. Above the second limit,
for each value of v3 there are two values of rp, and the quantum defect
surface is two lines. Finally, if we consider the energy region above the third
limit of Fig. 20.2, all three channels areopen for allvaluesof r. Inthiscasewe
canput abox of anyr > rc about the ionic core and inquire as to the normal
modes for electron scattering from the contents of the box. We have in fact
already solved this problem. The normal modes are the ^a wavefunctions.
Since the energy is above all the ionization limits we no longer need to
ignorethe r�> o� boundary conditions;they play no role. Geometrical
interpretation of the quantum defect surface In anyorbit of the Rydberg
electron, most of thetimeis spent when the electron is far from the nucleus,
where the wavefunction ismost reasonably characterized in terms of Ah the
coefficients of the (p( wavefunctions. Correspondingly, many of the
properties depend upon the values of At in a very direct way. As shown by
Cooke and Cromer,3 aparticularly attractive feature of QDT isthat the values
of A?, i.e. the composition of the wavefunction in terms of the collision
channels, can be determined by inspecting the quantum defect surface. If we
define the cofactor matrix Cia of the matrix of Eq. (20.12) by Cia = Cofactor
| Uia sin jr(vf + /ia)|, (20.14) then we can rewrite the determinant of Eq.
(20.12) as

422 Rydberg atoms Uia sin n(Vi + fia)Cia = 0 (20.15a) or >

t/,asinjr(v,+/OQ� = 0. (20.15b) / Comparing Eqs. (20.15a) and (20.15b) to
Eqs. (20.10a) and (20.11a),wesee that Cia = GAiBa, (20.16) where G is a
constant. Consider now the function /(viv2 � � � v,.) = det (Uia sin n(yt +
fia% (20.17) the same determinant set equal to zero in Eq. (20.12) to define
the allowed values of vt. If we construct the gradient V/we find the normal to
the surface of Eq. (20.12). Writing out the components,3 df d L d t | ^ i ( + O |
= � > U ja Sin Jt(Vj + IIa )Cja J n Uja COS n (v> = jiGAf. (20.18) If the
gradient defined by Eq. (20.18) isrepresented by avector perpendicular to the
quantum defect surface, its projections on the v{axes are proportional to the
values of Aj. In a two limit problem Eq. (20.18) reduces to the simple form8
A2 �dv 1 A\~ dv2 (20.19) Normalization The scattering/and
gwavefunctions we haveused arenormalized perunit energy, and now we
consider how to normalize the wavefunctions based on them in different
energy regions. First we consider the bound states. We require a bound state
wavefunction to satisfy 3 = l . (20.20)

Energy constraints 423 To do this we must change the 0, wavefunctions from

normalization per unit energy tonormalization per state, i.e. Eq.
(20.20).Usingthe derivative dWjVdv,- = Vvf we may convert the squared
wavefunctions |0,-|2 from energy to state normalization by multiplying by
\lv3. Equivalently, a bound 0, wavefunction which isnormalized per unit
energy has a normalization integral of v3. Since the wavefunction ^ = 2^0, is
composed of bound wavefunctions normalized per unit energy, its
normalization integral, N2, is given by 2 vl (20.21) inwhich the higher
vtstates carry proportionally more weight than dothe lowervt states. As
pointed out by Cooke and Cromer,3 thisweighting arises from the fact that the
At values reflect the fraction of orbits the Rydberg electron spends in the
various ^ wavefunctions of the state M*, and the factor v(3 reflects the time
duration of an orbit of effective quantum number vt. Using the normalization
integral of Eq. (21), the properly normalized bound wavefunction is given by
YAi%, (20.22) i inwhichthe ^ wavefunctions are stillnormalized per unit
energy. If we wished to expressEq. (20.22)interms of conventional
boundwavefunctions normalized per state, ^ B , using % = vx312 *,B, Eq.
(20.21) becomes (20.23) If one or more of the channels is open, the
wavefunction is a continuum wavefunction, since itextendsto r=
o�,anditmustbenormalized perunit energy. Each of the ^ continuum
wavefunctions is separately normalized perunit energy, so we simply require
for each p solution =1> (20-24) where the sum extends over the open
channels. Usingthe normalization relations of Eqs. (20.21) and (20.24)
andthe geometric relation between the At values we are able to construct
properly normalized wavefunctions at any energy. Energy constraints As
pointed out early in this chapter, 7ivtis really the phase of the / channel
wavefunction asr-^ o�. For each bound channel, wehave already introduced
the constraint Wt = � l/2vf which setsthe phase nvt for anyenergy Wt. If
weconsider,

424 Rydberg atoms for example, the region below the first ionization limit of
Fig. 20.1, all three channels are closed, and the states have discrete energies
and discrete valuesof v1, v2, and v3. Yet the quantum defect surface, defined
by Eq. (20.12), is continuous over v1? v2, and v3. More values of the phase
shifts Jivt satisfy Eq. (20.12) than are physically observed. What constrains
the continuous variation in vt allowed by Eq. (20.12) is the energy relation
W=Wi + Ii9 (20.25) which ties together the vl-values of allclosed channels.
Below the lowest ionization limit of Fig. 20.1, Eq. (20.25) isgiven by "-sH-
ii^-si**- <20 -26> a line, and its intersection with the quantum defect surface
defined by Eq. (20.12) yields a set ofpoints corresponding to the bound
atomic states. Between thefirst and second limits only channels 2and 3 are
closed, and Eq. (20.25) becomes W = �^ + I2 = �^ + /3, (20.27) 2vi 2vi 3
V ' which defines asurface independent of vx. The intersection of thissurface
with the quantum defect surface defines a line, and in this case all values of
v2 and v3 are sampled, although not all pairs v2and v3. The freedom ofthe
continuum phaseto take any value at any energy allows all values of v2 and
v3 tobe sampled. Alternative R matrix form of QDT The use of QDT which
has drawn ourattention to it is to represent autoionizing states. Consider the
simplest case of aseries of autoionizing states degenerate with a single
continuum, a two channel QDT problem. If we usetheICE method to observe
the autoionizing states we observe their positions and widths, which may be
characterized by two parameters, a quantum defect 6 and a scaled width n3F.
The relation of these quantities to the Uia and jua parameters of QDTisnot
obvious. In fact it is not even unique. In the two channel problem we are
considering there aretwo measured parameters, butthere are three parameters
required to specify the two channel QDT, jnl9 ju2, and a rotation angle 6to
specify Uia. The continuum phase is a superfluous piece of information for
the interpretation of an ICE experiment. If inour twochannel problem the
autoionizing states are channel 2 and the continuum is channel 1,the
wavefunction is given by V = A1<p1 + A2<p2, (20.28) where A1 = 1. The
spectral density ofthe autoionizing state isgiven byA2, which according to
Eq. (20.19) is given by the derivative dv1/dv2 since Aj = 1. The

Alternative R matrixform of QDT 425 position and widths of the autoionizing

states of channel 2 are determined by the derivative of the continuum phase
but do not depend at all upon the absolute continuum phase. Thus if we used
adifferent set of Uia and/ia parameters with the same value of dv1/3v2 at
each value of v2, these parameters would represent the series of autoionizing
states equally well, if only the positions and widths are observed. To arrive
at a more unique way of characterizing autoionizing states in the absence of
information about the continuum phase we can recast the QDT equations into
an R matrix form which is at the same time similar to the original
development of QDT from multichannel scattering theory.2The relation
between the different forms of scattering matrix is discussed by Mott and
Massey,11 Seaton,2 and Fano and Rau.12 If we rewrite Eq. (20.12)
usingsin(A + B) =sin A cosB + cos A sin B, it isgiven by cos (jijua) y [Uia
tanjivt + Uia tan jzjua] cos(jtv^Ai = 0. (20.29) i Since cos Jijua multiplies
the entire left-hand side of Eq. (20.29) we can ignore it and write the
remaining equation as a matrix equation. Explicitly [UT tan pv + tan pfi UT]
cos pv A = 0 (20.30) where tan TTV, tan TT/JL, and cos TTV are diagonal
matrices, and A is a vector. If we define the matrix R by R = UtanUT, (20.31)
and the vector a by a{ = cosjrvi^i, (20.32) then Eq. (20.30) can be rewritten
as (R + tan TTV)SL = 0. (20.33) The R matrix, or reactance matrix, is real
and symmetric. If there is no coupling between the / channels, off diagonal
elements of the R matrix vanish and the diagonal elements are given by
tanjrd,, where 6t is the quantum defect of the /th channel. To a first
approximation the interchannel couplings are given by the off diagonal
elements of the R matrix. In the form of Eq. (20.33) the R matrix both
accounts for the quantum defects of isolated channels and the interchannel
couplings. As pointed out by Cooke and Cromer,3 these are two rather
different functions. The quantum defect reflects only the phase shift of the
nonhydrogenic coulomb wave, and is a spherical, single channel effect. The
energy, angular momentum, and parity of the ionic core and Rydberg electron
are thus separately conserved. The interchannel couplings, on the other hand,
destroy the separate conservation of the above three quantities. We may
remove the quantum defects of the / channels from the R matrix by using
phase shifted coulomb waves asbasis

426 Rydberg atoms functions instead of the regular and irregular / and g
waves. We express the collision channel wavefunctions of Eq. (20.3) as2"4
<t>, = \ lXif'(Wh �, r) cos nv,- +Xig'(Wh �,r)], (20.34) where /' and gr are
phase shifted coulomb functions given by /' (wit �, r) = f(Wh �, r) cos n dt
- g(Wh �, r) sin jrdf- (20.35a) and g'(Wh �, r) = /(W,, �, r) sinjr<5, +
g(Wh �, r) cosJtdt (20.35b) in which v[ = Vi+ di. (20.36) If we define 4 =
At cos nvv (20.37) we can derive from Eq. (20.29) an expression analogous
to Eq. (20.33). Specifically, (R' + tan TTi/jQa' = 0, (20.38) where R' = [cos
TTS + R sin TTS]'1 [R COS TTS - sin TTS]. (20.39) If there are no
interchannel couplings, R' = 0, corresponding to the fact that the / channels
have quantum defects of dt and occur when v{ = 0. In general the R' matrix is
a real symmetric matrix with zeroes on the diagonal. Eqs. (20.33) and
(20.38) are precisely analogous to Eq. (20.12), and we can find the quantum
defect surface from Eqs. (20.33) and (20.38), for example, by setting det [R
+ tan TTV] = 0 (20.40a) or det [R' + tan TTV'] = 0. (20.40b) Another useful
property of the R matrix formulation is that we can write it in blocks
corresponding to the bound (closed) and continuum (open) channels.
Explicitly, we can write Eq. (20.33) as3 Rbb + tan m^ Rbc 1 ab = Q (20 41)
Rcb Rcc + tan TTTJ ac Eq. (20.38) may be written in exactly the same form.
In Eq. (20.41) we have replaced all the continuum phases by nx to reflect the
fact that we are searching for

The role of QDT All the normal scattering modes which satisfy Eq. (20.40a).
Following Cooke and Cromer3 we can write Eq. (20.41) in separate bound
and continuum matrix equations. Explicitly -[R + tan 7Tv\^Khc2ic = ab
(20.42a) and [Rcb[R + tanjTv\^Rbc -Rcc]ac = [tan 7rro]ac (20.42b) where
rp isone of the allowed values of x. From Eq. (20.42b) it is apparent that the
eigenvector ac, i.e. that linear combination of / channels, which satisfies Eq.
(20.42b) has a diagonal continuum-continuum R matrix given by the left-hand
side of Eq. (20.42b). In other words, it is the normal scattering mode. When
we solved Eq. (20.12) with some open channels we simply replaced nvx by
nx for all the open channels. Eq. (20.42b) justifies this procedure. The role
of QDT QDT provides a framework which relates a few energy independent
parameters to awealth of spectroscopic data. It isused both as an efficient
way to parametrize data and as away of comparing theoretical results to
experimental data. Which of the several parametrizations to use is usually
unimportant for comparing theoretical results to experimental observations,
since all the parametrizations are equivalent. On the other hand, if a set of
data is to be represented by QDT parameters, it isuseful to use the set of
parameters which allowsthe experimental data to be fit with the minimum
number of free parameters. For example, to fit ICE data, the phase shifted R
matrix approach is by far the most convenient, for the absolute continuum
phase does not enter. The QDT parameters can also be generated
theoretically, by ab initio13 or semiempirical methods.14~18TheR matrix
approach is averysuccessful example of the latter. In the application of the R
matrix method to Sr, for example, a model spherical potential is constructed
which correctly reproduces the energies of Sr+. The Schroedinger equation
for both valence electrons is then solved using this radial potential. The
Schroedinger equation is not solved over all space but onlyin a spherical box
of radius r0 using a truncated basis set of trial wavefunctions. The size of the
box is larger than the wavefunction of the highest energy state of Sr+
considered in the calculation. Thisrestriction sets an upper limit on the
energiesof thestateswhich canbetreated in thisway, but
makessolvingthecoupled equations a fairly straightforward matter. The
restriction also ensures that the wavefunction at the edges of the box is a one
electron wavefunction. The Schroedinger equation is solved for the normal
scattering modes, and the logarithmic derivative of the solutions at the box is
related to the scattering phase shift Jtjua. The results can be expressed as the
phase shifts and the matrix which connects the

428 Rydberg atoms wavefunctions of the normal modesto, for example,the LS

coupled basis,or an R matrix can be specified for the LS coupled basis.
References 1. M. J. Seaton, Proc. Phys.Soc. London 88, 801 (1966). 2. M. J.
Seaton, Rep. Prog. Phys. 46,167 (1983). 3. W. E. Cooke and C. L. Cromer,
Phys.Rev. A 32,2725 (1985). 4. A. Giusti, /. Phys. B 13, 3867(1980). 5. A.
Giusti-Suzor and U. Fano, /. Phys. B 17,215 (1984). 6. K. Ueda, Phys.Rev. A
35,2484(1987). 7. J. P. Connerade, A. M. Lane, and M. A. Baig, J. Phys.B
18,3507(1985). 8. U. Fano, Phys.Rev. A 2, 353 (1970). 9. K. T. Lu and U.
Fano, Phys. Rev. A 2, 81 (1970). 10. C. M. Lee and K. T. Lu, Phys.Rev. A 8,
1241 (1975). 11. N. F. Mott and H. S. W. Massey, TheTheory ofAtomic
Collisions (Oxford University Press, New York, 1965). 12. U. Fano and A.
R. P. Rau, Atomic Collisions and Spectra (Academic Press, Orlando, 1986).
13. W. R. Johnson, K. T. Cheng, K. N. Huang, and M. LeDourneuf, Phys.Rev.
A 22, 989 (1980). 14. C. H. Greene and L. Kim, Phys.Rev. A 36, 2706
(1987). 15. C. H. Greene, Phys.Rev. A 28,2209 (1983). 16. M. Aymar, E.
Lue-Koenig, and S. Watanabe, /. Phys. B 20, 4325 (1987). 17. M. Aymar and
J. M. Lecompte, /. Phys. B 22, 223 (1989). 18. C. H. Greene and Ch. Jungen,
inAdvances in Atomic and Molecular Physics, Vol. 21, eds. D. Bates and B.
Bederson (Academic Press, Orlando, 1985).

21 Optical spectra of autoionizing Rydberg states QDT enables us to

characterize series of autoionizing states in a consistent way and to describe
howthey are manifested in optical spectra. We shallfirstconsider the simple
case of a single channel of autoionizing states degenerate with a continuum.
Of particular interest is the relation of the spectral density of the autoionizing
states to how they are manifested in optical spectra from the ground state and
from bound Rydberg states using isolated core excitation. We then consider
the case in which there are two interacting series of autoionizing states,
converging to two different limits, coupled to the same continuum. First we
consider the two channel problem shown in Fig. 21.1. Our present interest is
in the region above limit 1, i.e. the autoionizing states of channel 2. Later we
shall consider the similarity of the interactions above and below the limit. A
typical quantum defect surface obtained from Eq. (20.12) or (20.40) for all
energies below the second limit is shown in Fig. 21.2. The surface of Fig.
21.2 may be obtained with either of two sets of parameters, dx = 0.56, 52 =
0.53, and R'l2 = 0.305, R'n = R'22 = Q orfix = 0A,ju2 = 0.6, and Un = U22 =
cosO and U12 = � U2i =sind, with 0 =0.6 rad.12To conform to the usual
convention, inFig. 21.2 the Vj axisis inverted. The wavefunction is given in
terms of the collision channels by V = A1</>1+A24>2. (21.1) Between limit
1 and limit 2, since 0X is open, A\ = 1. If we choose At = 1, it is straight
forward to show that Fig. 21.1 Two atomic channels associated with two
states of the ion (ionization limits), shown bybold lines. Above thefirstlimit
and below the second the nominally bound states of channel 2autoionize into
the continuum of channel 1. 429

430 Rydberg atoms Fig. 21.2Quantum defect surface for the twochannel
problem showing vx, or equivalently r, the phase shift in the open channel,
channel 1,divided by �jr. The vx axisisreversed to conform to convention.
The probability offindingan autoionizing state of channel 2 at anyenergy is
given byA2, which we shall term the spectral density. Squaring Eq. (21.2)
we find A2 From Eq. (21.3) several points are apparent. FirstA2 is
periodicin v2. Second the maximum value, A* = 1/(R[2)2, occursfor v2 = 0
(mod 1), and the minimumvalue ^2 = (#i2)2, occurs for v2 = 0.5(mod 1). For
\R[2\ � 1 Eq. (21.3) can be approximated by a Lorentzian, i.e.1 from which
it is apparent that the half maximum points of A2 occur at 71
)2/ThustheFWHMisgivenby2(R'12)2/JV or in terms of the energy width1 .
(21.6) Eq. (21.6) showsboth the expected v"3 scaling of the autoionization
rates and the utility of the parametrization in terms of du 62, and i^It isuseful
to recall the geometric interpretation of the quantum defect surface of Fig.
21.2.Aswe have already discussed, we can drawthenormal to the curve of
Fig.21.2, anditsprojections onthevx and v2 axes give therelativevaluesofA%
and A%. In Fig. 21.2 the normal points nearly vertically, i.e. in the vx
direction

Optical spectra of autoionizing Rydberg states 431 20 Fig. 21.3The spectral

density A\ of the autoionizing state vsv2. except at v2 = 0.4, where it has
asubstantial horizontal, or v2, component as well. In other words, only at
v2�0.4 is there an appreciable value of A�, and here is where the
autoionizing state of channel two is located. One might ask why we bothered
with the normal tothe quantum defect surface, sincefor thetwochannel case
Eq. (20.19) shows that A\lA\ is given simply by -dv1/dv2. The reason is that
the normal can be generalized to three dimensions. In any case, we plotA\
inFig. 21.3, and it is clear that all the information contained in Eq. (21.3) is
in the A% curve. The autoionizing states of channel 2 are represented by
Fig.21.3, however they donot necessarily appear as inFig. 21.3 in a
photoionization spectrum. Let us first consider photoexciting the autoionizing
states of channel 2 and the degenerate continuum of channel 1 from a compact
initial state, g, such as the ground state. Since the initial state isspatially
localized near the ionic core, only the part of the Rydberg wavefunction near
the core plays an active roleinthe excitation. We can write the dipole matrix
element for the excitation in either of twoways,1"3 > � Bada (21.7a) a or ]T
t dt cosJt{vt + 0f). (21.7b) In Eqs. (21.7) da and dx are energy independent
dipole matrix element constants defined by da = <*>|*a> (21.8a) and

432 Rydberg atoms Fig.21.4 Thevalue of the squared a channel coefficient

B\. Apart from a factor of w it is proportional to thephotoionization cross
section, if da=1 = 0 as assumed. dt cos7t(Vi + 4>i) = <�>|0i>. (21.8b)
Using the a channels takes advantage of the fact that near the origin the phase
of the a channels is energy invariant. Furthermore, it is often the case that
only one of the da is nonzero. For example, if the a channels are LS coupled
and the initial state g is a singlet state, da for the triplet a channel vanishes.
Writing the excitation matrix element intermsof the/channels, we must
account for thephase variation of the vx channels with energy. Although such
excitations can be described usingthe /channels,most often they are described
usingthe achannels, because it is simpler to do so. In either case, the
photoionization cross section is given by a = ^ | < � > | � > | 2 . (21.9) As a
typical example, wecompute (W^ju^) assuming that da =0 for a = 1 and da= 1
for a = 2.UsingEq. (21.2)we calculateA2, assumingAx = l,andusingEq.
(20.11a) wecalculate B2 (we do not need B1 since da=1 =0). In Fig. 21.4 we
show the squared dipole moment, B\, which exhibits the asymmetric Beutler-
Fano lineshape characteristic of autoionizing states.4Theselineshapes
areoften characterized bythe Fano q parameter which is the ratio of the
matrix elements per unit energy connecting the initial state to the discrete
autoionizing state and to the continuum. When q > 0 the asymmetry isasshown
in Fig. 21.4, and when q < 0 it is reversed; the zero in the cross section is on
the high energy side of the autoionizing state. When \q\ � 1the continuum
excitation is negligible and the Beutler-Fano profile becomes a symmetric
Lorentzian, and when q = 0 the Beutler-Fano profile is a symmetric
Lorentzian dip in the cross section. When \q\ ~ 1 the lineshape is most
asymmetric.

Optical spectra of autoionizing Rydberg states 433 If we again consider Fig.

21.4,wecan see that the crosssection vanishes at v2 = 0.32 and that the
profile does not match the spectral density, A2, of the autoionizing state. The
Beutler-Fano profile of Fig. 21.4 is periodic in v2 with period 1, so the
spectrum from the ground state consists of a series of BeutlerFano profiles.
At higher values of v2 the profiles become compressed in energy since
dW/dv2 = l/v2. Fig. 19.2 shows two regular series of Beutler-Fano profiles
between the Ba+ 6p1/2 and 6p3/2limits. In this case the absorption never
vanishes because there is more than one continuum. Now let us consider the
ICE of the autoionizing states of channel 2. For example, imagine that we are
able to start from a bound Ba 6sn� Rydberg state and drive the transition to
the autoionizing 6pn� state of channel 2. Channel1 is the degenerate
continuum. For the moment we ignore the spins of the electrons. The 6s
electron makesthe dipoletransition tothe 6p state and theouter electronis
projected onto the autoionizing nl state. For the reasons given in Chapter 19
we can ignore the amplitude to the continuum. For concreteness, we define
(j)\ = continuum, 02 = 6prc�, (21.10) % = 6wi�b. The wavefunction ty of
the autoionizing state is given by Eq. (21.1), and to calculate the
photoionization cross section we need the dipole matrix element =
<<*>b|//|A202)- We can write (>Pb| and |02) as product wave functions;
(21.11) and |02> = |Il2)|6p)|v2B�b), (21.12) where (6s| and |6p) are the
ionic wavefunctions, (vb�B| and |v2�) are radial wavefunctions for states
of effective quantum numbers vb and v2, and (fib| and \ft2) are angular
wavefunctions. The superscript Bin |vb�B) reflects the fact that it is
normalized per unit state,while |v2t) is normalized per unit energy.
Substituting the expressions of Eq. (21.12) into Eq. (21.11) and computing
the squared matrix element leads to (21.13) In Eq. (21.13) (6s[u|6p) is just
the Ba+ 6s-6p dipole moment, and dUh is the Kronecker delta function. Since
�b=� we shall simply write �in place of �b. All the energy dependence
in the cross section arises from the factors A2 and Kvb�B|v2�)|2. These
are, respectively, the spectral density of the channel 2 autoionizing states, and
the overlapintegralbetween thebound andcontinuum nt states with effective
quantum numbers vb and v2. We have already seen that A2 is simply given
by �dv1ldv2, the derivative of the quantum defect surface, and repeats
modulo 1 in v2. The overlap integral is given is closed form by5

434 Rydberg atoms ( 2 1 M ) ^ K - v2) (vb + i>2) This expression is valid

for any � state, as long as � � n. The factor i>2/2 accounts for the fact that
|v2�) is normalized per unit energy. The overlap integral (vb�B|v2�B)
between two bound state wavefunctions would not have the factor v2/2 and
would be equal to one for v2 = vb and zero for v2 different from vb by an
integer, as expected. We can write the cross section as o = ^A22\
(vbdB\v2d)\2. (21.15) c In Fig. 21.5we plot A2, |(^bdB|v2d)|2, and the
resulting cross section for the casein which the initial state has an effective
quantum number vb = 12.35 and the autoionizing states of channel 2 are
located at v2 = 0.28 (mod 1) and have fractional widths v2F = O.I.6 In this
case the central lobe of the overlap integral coincides with the autoionizing
state peaked at v2 = 12.28, and there is one strong peak in the cross section,
at v2 = 12.28. There are also much weaker subsidiary peaks but for the
moment weignore them, our present interest being to verify that we can
calculate the basic features of the ICE spectra. Since the width of the central
lobe of the overlap integral is much larger than A2, the width of the cross
section is effectively equal to A2? as asserted in the discussion of ICE in
Chapter 19. It is interesting to consider what the spectrum would look like if
the autoionizing states of the same width were located at v2 = 0.85(mod
1)instead of 0.25(mod 1), i.e. if the quantum defects of the bound and
autoionizing states differed by 1/2. In this case the spectrum is given by Fig.
21.6.6 The central lobe of the overlap integral contains now the two
autoionizing states at v2 = vb� 1/2, so in this casewe expect two strong
peaks in the ICE spectrum with noticeable subsidiary shake up, Av # 0,
peaks. In this case the peaks of the cross section do not match the variation
inA2 sowell since they occur on the sloping sides of the central lobe of the
overlap integral. The severity of the distortion depends upon how wide the
autoionizing states are. It is most often the case that the quantum defects of the
bound and autoionizing states do not differ by more than 0.2, in which case
the overlap integral is constant across A2 and the ICE spectrum reflects A2
accurately. The shake upsatellites If we examine Eq. (21.14) and Fig. 21.5 it
is apparent that apart from v2 = vb, the cross section has zeroes when v2 =
vb(mod 1). Furthermore, the combination of these zeroes and the periodic
variations inA2 lead to satellite structure whichis

The shake up satellites 435 EFFECTIVE QUANTUM NUMBER v 9.82

10.59 11.57 12.89 14.79 17.89 1 CM CM cc .a 440 Q CC LU O 62,850
63,009 63,168 63,327 63,486 63,644 TERM ENERGY (cm"1) Fig. 21.5
Plots of (a) the spectral density, (b) the square of the overlap integral
between initial and final states, and (c) their product which is proportional to
the optical cross section o. All are calculated for the third laser excitation
from the 6sl5d (^ = 12.35) state to the (6p3/2ftd)j = 3 channel (from ref. 6).

436 Rydberg atoms 4400 9.82 10.59 11.57 12.89 14.79 17.89 EFFECTIVE
QUANTUM NUMBER v2 Fig. 21.6 Calculated cross section o from a bound
state of effective quantum number vh = 12.35 toan autoionizing Rydberg
series with width T = O.llv"3 and quantum defect d2= 0.15 so that the
autoionizing states are found at v2 =0.85(mod 1) (from ref. 6). very sensitive
to the difference between 6b and d2, as shown by Figs. 21.5 and 21.6. In
particular it isvery easy to see small changes in d2 � db, if they are almost
equal. Although the structure away from v2 = vh is weak, especially in Fig.
21.5, itis possible to observe it with high laser power and compare it to the
spectra calculated using Eq. (21.13). The high power saturates the center of
the spectrum where v2�v b, and to account for the saturation the ion signal
must be expressed as7'8 7= /0 (l-e-^) (21.16) where O is the time integrated
photonfluxper unit area. Tran et al.9 have excited Ba atoms in a thermal beam
to the 6sl5d *D2 state with two circularly polarized lasers. They then
scanned the wavelength of a third circularly polarized laser in the vicinity of
the Ba+ 6s1/2 �> 6p3/2 transition and detected the ions resulting from
theexcitation of the 6p3/2nd / = 3 states. Their observed 6sl5d 1 Dz-
6p3y2wd / = 3spectrum is shown inFig. 21.7. The two sharp peaks at 21850
and 22170 cm"1 are two photon resonances and should be disregarded. From
the 6snd *D2 states only asingle 6p3/2nd 7 = 3 series is excited, so for all
practical purposes this is a two channel problem, a single series of
autoionizing states and the degenerate continuum. The presence of many
continua is irrelevant as long as the excitation is insensitive to the continuum
phase. In Fig. 21.7 the broken curve is calculated using Eqs. (21.15) and
(21.16) with oQ ~ 100 atthe peak ofthe cross section, and the agreement
between the experimental and theoretical spectra is excellent. Itis also
interesting to note that the zeroes are still evident even though the signal at
the peak cross section issaturated by a factor of 100.

The shake up satellites 437 EFFECTIVE QUANTUM NUMBER v2 9.82

10.59 11.57 12.89 14.79 17.89 < z CO z o I I I I I _BaII6p1/2T-6p3/2nd f
_BaII6s1/2/ 7l5s15d - / j 6s6p / 14d -J_6s2 * - 13d j \ , I \ \ / V l / I i I I I 16d
_ 17d \ I 18d 21535.0 21693.8 21852.6 22011.4 22170.2 22329.0 PHOTON
ENERGY ftco (cm"1) Fig. 21.7Photoexcitation spectrum of the 6sl5d *D2
�> (6p3/2wd)/ = 3 transition inBa as a function of the frequency of the third
laser. All three lasers are circularly polarized in the same sense. The broken
line isthe spectrum calculated from Eqs. (21.15) and (21.16). The energy
level inset is not to scale (from ref. 9). We have sofarconsidered the
excitation to two channel systems, whereas in reality most ICE experiments
have been done with autoionizing states converging to the lowest p states of
alkaline earth ions, as shown schematically in Fig. 21.7. We now wish
toconsider exciting the Ba 6pynd state ofhigh n and following the cross
section across the 6p; limit. To describe this cross section requires that we
calculate the overlap integral tothe continuum as well as tobound
autoionizing states. In the explicit expression for (vbdB\v2d) of Eq. (21.14)
we can see that the integral iszero when v2 = Vb(mod 1), by orthogonality,
and maximum forv2 = vb+l/2(mod 1). The same reasoning can be extended to
thecontinuum. If the phase Jtv2 is such that v2=vb the overlap integral
vanishes by orthogonality, and if v2 = Vb + l/2(mod 1) it is a maximum. In
fact, forany continuum phase nv2 the overlap integral is the extrapolation of
the bound (vbtB\v2�) integral for the same value of r2(mod 1)- In Fig. 21.8
we show overlap integrals from the bound Ba 6sl2d X D2 and 6s23d l T>2
states to both the 6p1/2rcd and 6p3/2nd channels.6The overlap integrals in
the continuum are shown for the phases jt(db�l/2) which give the maximum
absolute values of the overlap integral. The phase dependent continuum
excitation has been observed by Tran et al.6 and more clearly by Story and
Cooke.10 They excited the Ba 6sl9d X D2 states to the region ofthe 6p1/2
limit and observed the fluorescence from the excited 6p1/2 state of Ba+. The
resulting spectrum is shown inFig. 21.9. Using this technique they do not
detect excitation below the 6p1/2 limit atall efficiently, but are ableto see the
variations in the cross section above the 6p1/2 limit with remarkable clarity.
438 Rydberg atoms 60,000 61,000 62,000 63,000 64,000 65,000 TERM
ENERGY (cm"1) 1 - <b) 1 x 10 I UPI/2 I I x 10. H i l > \ \ / / f I : 10 6P3/2
100 50 0 -50 -100 60,000 61,000 62,000 63,000 64,000 65,000 TERM
ENERGY (cm"1) Fig. 21.8 Overlap integrals (vhdB \ v2d) from (a)theBa
6sl2d state (vh = 9.36) and (b) the 6s23d state (vb = 20.29). In (a)
theintegrals from 6sl2d to the6p1/2nd channel ( ) and 6p3/2nd channel (�)
are multiplied by 10at 62,000 and 63,800 cm as indicated. Note that the
integrals extend smoothly across the 6p179 and 6p3/2 limits. Note also that
the overlap integrals peak at -61,000 and -62,800 cm"^ near the 6p1/212d
and 6p$/212d states. The maximum value is (9.36)3/2, not 1, because of the
continuum normalization. In (b) the integrals from 6s23d to the 6p1/2nd
channel ( ) and 6p3/2nd channel (�) are each multiplied by 10inthe energy
ranges far from their peak values at 62,000 and 63,800 cm"1, respectively. In
the continua above the 6p1/2 and 6p3/2 limits the maximum positive and
negative values of the overlap integral areplotted (from ref. 6). Since the
features due to the 6p3/2nd states decrease in intensity with increasing n, it is
evident that most of the excitation is due to the 6p1/2�d continuum, not to
the 6p3/2nd states. As shown by Fig. 21.8, in Ba the overlap integrals for the
6p1/2nd and 6p3/2nd channels overlap, admitting the possibility of
interference in the excitation amplitudes to the 6p1/2nd and 6p3/2nd
channels. However the fine structure

The shake upsatellites 439 10r8 if) 1 6 1 4 o c �CO 0 I , � � . I , . , , 1

20500 21000 21500 Energy (cm"1) Fig. 21.9 Shake-off spectrum of the
transition, Ba 6sl9d �> Ba 6p1/2�d, obtained by detecting the fluorescence
from the excited Ba+ 6p1/2 ions. The resonances are Ba 6p3/2�d states
with 11 < /i < 14 (from ref. 10). splitting of the Ba 6pj states is large enough
that the interference is not unambiguouslydetectable.
Belowthe6p1/2limititisdifficult toseesmall interference effects due to the
6p3/2ndexcitation amplitude in the presence of the enormous effects of the
interseries interaction. Above the 6p1/2 limit the excitation amplitude to the
6pi/2�d continuum appears much the same as the excitation of the 6p3/2ftd
statesas shown in Fig. 21.9. The situation in Sr is very much the same.
However, in the lighter alkaline earth atoms Ca and Mg the effect of
interference in the two excitation amplitudes is very apparent and must
beproperly taken into account to reproduce the observed spectra.1112 The
presence of the overlap integral in the ICE cross section appears at first
glance to be at best a minor inconvenience, certainly not something which is
useful. However, the presence of the overlap integral variation has several
uses. Sandner etalP have used the overlap integral to determine the effective
quantum number of the initial bound Ba 6s24d l T>2state to one part in 104.
Combining the implied bindingenergywiththeterm energyoftheinitial Rydberg
stateprovides a novel way of measuring the ionization limit. The overlap
integral also provides a way of measuring small admixtures of different
states. For example, Kachru etal. have determined that there isa 2%
admixture of the 6p3/2ndj state in the6p3/2ns1/2 state and vice versa.14 This
admixture is so small that it would normally be quite difficult to detect,
especially in an autoionizing state. However, the overlap integral variation
allows it to be observed clearly.

440 Rydberg atoms Interacting autoionizing series We would now like to

consider the more complex case of two series of autoionizing states
interacting with each other and with a degenerate continuum. The Ba 6pn�
states just below the Ba+ 6p1/2 limit constitute a good example. As
shownbyFig. 19.2,inthevuvspectrum from the ground state,just belowthe
6p1/2 limit the spectrum has structure due to the series converging to both the
6p1/2and 6p3/2 limits. How much of the structure is due to the interseries
interaction and howmuchtoBeutler-Fano interference profiles is not clear.
Unfortunately, there is no purely experimental way to tell. In contrast, the ICE
method allows us to differentiate between interference in the excitation
amplitudes and the effects of interseries interactions. As afirst example
wewould like to consider the Ba 6pnd / = 3 seriesjust belowthe Ba+
6p1/2limit,the region shown inFig.21.10. We shall treat this problem as a
three channel quantum defect theory problem with the three channels 0! = / =
3continuum, 02 = 6pi/2nd5/2 J=3, 03==6p3/2nd; J = 3. (21.17) In Fig. 21.10
we show the 6pnd states as being degenerate with only a single continuum,
above the 6s1/2limit.15 This simplifying assumption allows us to treat the
problem with only three channels. It is important to recall that in recording an
ICE spectrum we do not directly excite the continuum, it acts only as a sink
for electrons. As a result, the fact that the continuum is not well characterized
is not important. Above the 6p1/2limit the Ba 6p3/2nd/ = 3 states accessible
from the bound 6snd l T>2 states and the linear combination of continua into
which they autoionize can be treated as a two channel problem. Below the
6p1/2 limit we must use the three channels of Eq. (21.17). In the region
below the Ba+ 6p1/2 limit the / = 3 wavefunction is given by V = Arfi +
A2<p2 + A303, (21.18) where A1 = 1. An interesting point to note about Fig.
21.10 is that the 6p3/210d state isdegenerate with the 6p1/2nd5/2 states of n
~ 20. If we observe the spectrum from the 6s20d X D2state to the
6p1/220d5/2 state we observe a single Lorentzian line, as we expect for ICE.
If, on the other hand, we examine the 6sl0d �> 6p3/210dj spectrum
wefindthe spectrum of Fig. 21.11, a broad envelope containing sharp
structure.16 The structure evidently comes from the 6p1/2nd5/2 states. At
first glance it seems evident that the observed spectrum of Fig. 21.11 is a
sequence of Beutler-Fano interference profiles, which reverse in the Fano q
parameter across the line. Although the solution may ultimately be expressed
in the same mathematical form,1 asimple consideration of the excitation
amplitudes

Interacting autoionizing series 441 64 000 63500 k ~ � 2 < 1 ^ : = 18 z 16

� 14 63 000 __ ~ 12 62 500I- Ba-6P l / 2 _ 'OIL 11 E 62 000 10 O en 2 61
500 61000 60 500 59 500 + 6p3/2 ////////p 1 / 2 5 / 2 20 = _ 18 16 14 12 11
10 60 000 8 9 Fig. 21.10 The 6prcd7 = 3levels of Ba converging to the 6p1/2
and 6p3/2 limits. Note that the 6p3/2 lOdj state is degenerate with the
6p1/220d5/2 state. The 6p3/2lldj and higher states are above the 6p1/2 limit.
All the 6pnd states lie in the continuum above the Ba+ 6s1/2 state (fromref.
15). indicates that the structure is not due to interfering amplitudes. There are
three energetically possible outcomes of the 6sl0d state absorbing aphoton;
transitions to the 6p3/210dj, 6p1/220d5/2, and 6s1/2rf states.15If we
represent the 6sl0d *D2 state by the wavefunction "9b and use the fact that it
has an effective quantum number vb= 7.3,we can write the dipole matrix
elements from the initial 6sl0d X D2 state to the three / channels of Eq.
(21.17) as (21.19) where (65r|r|6p) isthe Ba+ radial matrix element, and fl2
and Cl3 are angular factors of order one which account for the angular
momenta and spins of the two

442 Rydberg atoms 62095 61865 61635 WAVENUMBER (cm"1) Fig. 21.11
The experimental 6sl0d 1 D2-6p3/210dj ICE spectrum is shown by the bold
line ( )�The structure produced byinteraction with the 6p1/2nd5/2 states
isquite evident. A three channel, QDTfitis also shown by the light line ( )
(from ref. 16). electrons. In Eq. (21.19) the continuum excitation is
vanishingly small, due to the very different spatial variations of the vb and ei
wavefunctions. The overlap integral (vbdB | v3d) is an overlap integral
between two lOd wavefunctions, both having v ^ 7.3, and is equal to v3/2
while the overlap integral (vbdB | v2d) is an overlap integral between lOd
and 20d wavefunctions, and is thus much smaller than the channel3
overlapintegral (vbdB | v3d).In otherwords,onlythe amplitude to the
6p3/210dj state is important, and to a good approximation (21.20) In the last
expression of Eq. (21.20) the squared overlap integral is approximately
constant and equal to v\ for v3 ~ vb, and (6s|r|6p) and ft3 are constant aswell,
so the crosssection isdirectly proportional toA\. Knowing that the structure of
Fig. 21.11 has no contribution from interference in the excitation amplitudes
vastly simplifies the task of interpreting the spectrum. In fact, knowing that
all the information appearing in the spectrum isinA$, we know that it must be
found in the quantum defect surface. In Fig. 21.12 we show the quantum
defect surface for the three channel 6pnd J = 3 problem.16 This surface,
inscribed in a cube, is valid for all energies below the 6p1/2 limit. If there
were no interchannel interactions the surface of Fig. 21.12 would consist of
three intersecting planes at vx =0.2, v2 = 0.2, and v3 =0.3. The

Interacting autoionizing series 443 � v 3 Fig. 21.12 The three dimensional,

quantum defect surface for the 6pnd states below the 6 Pi/2 limit. Here v2
and v3 are the effective quantum numbers relative to the 6p1/2 and 6p3/2
limits, respectively. nvx is the continuum phase. The direction of the normal
to the surface at any point indicates the amounts of 6pi/2rcd5/2,6p3/2rcd/,
and continuum character in the wavefunction (from ref. 16). interactions
between the pairs of channels lead to the avoided crossings between the
planes. For example, the avoided crossings between the v1 =0.3 and v2 = 0.2
planes occurs because of the coupling between channels 1 and 2. Aswe have
described intwodimensions, constructing the normal tothe surface at any
point allows us determine the relative values of A? at that point; the
projections of the normal in the /directions are proportional to A?. For
example, in the region around the letter C the normal points in the vt
direction, so the wavefunction is predominantly channel 1, the continuum.
More precisely, A2=l, Al � 1 and Aj � 1. At letters A and B the normal
points in the v 2 direction sothe wavefunction is predominantly channel 2,
i.e. A2 = l,Al�l, and Aj � A%. At letter D the normal points in the v3
direction, and A2 = 1,A2 � 1, andAl � 1. In other words, simply inspecting
the quantum defect surface of Fig. 21.12 allows usto determine visually the
values of Af at any point on the surface. Nowweneed to determine which
points of the surface of Fig. 21.12 correspond to the energy scan of the
spectrum of Fig. 21.11. The path along the quantum defect surface is
determined by the energy constraint

444 Rydberg atoms 10 15 20 25 305 V 2 Fig. 21.13A plot of v3 vs v2. As

v3�> 8.058v2�> ��. For the Ba+ 6p; states I3 � I2 = 1690 cm"1, or
7.70 x 10r3 au. Expressing v3 as a function of v2, Eq. (21.21) becomes (/3 -
/2) +4j, (21.22) Eq. (21.22) defines a 0 independent two dimensional
surface. In Fig. 21.13 we plot v3 as a function of v2, i.e. the projection of the
surface onto the 0 = 0 plane, which shows that v3�> 8.058 as v2�> ��,
and that near the 6p1/2 limit v2 increases much more rapidly than does v3.
We can plot Fig. 21.13 modulo 1 in both v2 and v3 over the range of v2 and
v3 which corresponds to the energy scan of Fig. 21.11. Over the energy range
of Fig. 21.11 v2 varies from 12.89 to 23.33, and v3 varies from 6.82 to 7.61,
and in Fig. 21.14 we show v3(mod 1) vs v2(mod 1) over this range. In each
branch the curve reaches the right-hand side at v2 = 1 and some value of v3.
The intersection of the surface defined by Eq. (21.22) and shown in
projection in Fig. 21.14 with the surface of Fig. 21.12 gives the path along
the quantum defect surface corresponding to the spectrum of Fig. 21.11. The
next higher branch of the curve starts at v2 = 0 at the same value of v3.
Equivalently, the projection of the path along the quantum defect surface on
the v2v3 plane as v3 varies from 6.82 to 7.61, and v2 from 12.89 to 23.33, is
shown by Fig. 21.14. Before we see how the spectrum of Fig. 21.11emerges
from Fig. 21.12, let us try a simpler case, the 6sl6d X D2 �> 6p1/216d5/2
ICE transition, which exhibits a Lorentzian peak at v2 = 13.3.16 Since the
peak falls at the center of the (vbdB|v2d) overlap integral the cross section is
proportional to A2, which should thus have a maximum at v2 = 13.3.
Inspecting Fig. 21.14 we can see that the 6p1/216d5/2 state lies on the last
full branch of the v2, v3 curve, from v3 = 6.85 to v3 = 6.98 and v2 = 13to v2
= 14. This branch of the curve of Fig. 21.14 is almost parallel to the v2 axis
and lies at v3 ~ 6.9 where the normal of the quantum defect surface points in
the v1 direction except at v2 ~ 13.3 where it points predominantly in the v2
direction. In other words the normal to the quantum defect surface also tells
us that A2 is
Interacting autoionizing series 445 Fig. 21.14 A plotofv3(mod1) vs v2(mod
1) overthe range 6.82 < v3 < 7.62 and 12.89 < v2 < 23.33, corresponding to
the energy scanof Fig. 21.12. peaked at v2=13.3, the location of the
6p1/216d5/2 state. The projection of the normal on the v3 axis is minimal, so
A3 isvery small. Nowwereturn to the 6p3/210d spectrum of Fig. 21.11. It
extends from v3 = 6.82 to v3 = 7.61 and correspondingly v2 = 12.89-23.33.
From Fig. 21.14 we can see that the range v3 = 6.8 to v3 =7.6 corresponds to
ten branches of the vxv3curve of Fig. 21.14. Correspondingly, we follow the
quantum defect surface above these branchesinthe v2,v3plane. A3 istheratio
oftheprojections ofthenormalin the v3 and vx directions. Inspecting Fig.
21.12, it is apparent that for v3 ~ 7 there is almost no component in the v3
direction. As v3 increases to 7.2 the v3 component of the normal increases
uniformly except when v2 ~ 0.2(mod 1), the locations of the
6p1/2nd5/2states, and where the three planes of the quantum defect surface
intersect, when it drops sharply in favor of the normal in the v2 direction. In
other words, A^ has a broad maximum with holes from the 6p1/2nd5/2 states.
This description is in qualitative agreement with the observed spectrum of
Fig. 21.11. In Fig. 21.11 the light line shows the spectrum calculated using
the three channels of Eq. (21.17). Specifically, A3 is shown. It was obtained
from the expression of Eq. (20.16) for A{ with the requirement that A2 = 1.
Usingthe phase shifted Rmatrix approach Cooke and Cromerl have derived
an algebraic expression for the calculated spectrum of Fig. 21.11. In
problems analogous to this one Connerade17 has treated the broad state
analogous to the 6p3/210d state as a continuum of finite bandwidth into
which a series of states analogous to the 6p1/2nd5/2 states autoionize, and
Cooke and Cromer1have shown that the QDT expressions reduce to this
form. The three channel treatment leading to the calculated curve of Fig.
21.11 istoo simple to reproduce the experimental spectrum with great
accuracy. First, having only one continuum forces the theoretical spectrum to
have zeroes which are not present in the experimental spectrum. Second, we
have ignored one of the Ba

446 Rydberg atoms CO 4545 4560 4575 4590 WAVELENGTH (A) Fig.
21.15 (a) Observed Ba 6sl2s % -> 6p3/2l2s1/2 spectrum showing the
structure produced by the interacting 6p1/2ns1/2 and 6p1/2�d3/2 states, (b)
Calculated spectrum using a six channel, QDT model (from ref. 1). 6p3/2nd
7 = 3 series. While it is not visibly excited from the 6snd X D2 states, it is
slightly coupled to the 6p1/2nd5/2 states, and thus cannot properly be
ignored in an analysis of the spectrum. A more realistic model does a much
better job of representing the observed spectra. An elegant example is the
ICE spectrum from the bound 6sl2s % state to the 6pns J = 1 states just
below the 6p1/2 limit, the region containing the 6p3/212s1/2/ = 1 state and
6p1/2ns1/2 and 6p1/2nd3/2 states of n ~ 25. The experimental spectrum is
shown in Fig. 21.15.1 Two points are worth noting. First, this spectrum
corresponds to the range 61,996 cm"1 to 62,236 cm"1 in Fig. 19.2, i.e. the
uninterpretable region just below the Ba+ 6p1/2limit. Second, applying the
arguments used in connection with Eq. (21.20), we can see that the spectrum
of Fig. 21.15 reflects Af of the 6p3/2rcs1/2/ = 1 channel. In Fig. 21.15 we
also show the calculated spectrum based on a six channel QDT model, the
6p1/2ns1/2, 6p1/2rad3/2, and 6p3/2ns1/2J = 1 closed channels and three
continua. As shown, the agreement is superb. Thus far our discussion of the
interacting series has focused on the spectrum of the lower v state converging
to the higher 6p3/2limit. There are, however, several interesting aspects to
the 6p1/2 series as well. At first glance the 6sns �> 6p1/2ns1/2 spectra are
unremarkable; they are simple Lorentzian peaks. However, if we return to
our original consideration of interacting bound series, we would expect
significant energy shifts of the 6p1/2rcs1/2 states. In fact this is not the case,
as is shown by Fig. 21.16, a plot of the quantum defects of the 6p1/2ns1/2
states, which

Interacting autoionizing series 447 4.4 o LLJ LJL LJU Q 3 o 4.2 JX o s 1 G X

X G i 1 G BI i 4.0 18 22 26 30 QUANTUM NUMBER Fig. 21.16 Observed
(X) and calculated (�) quantum defects of the Ba 6p1/2ns1/2 states, which
are degenerate with the 6p3/212s1/2 state. Note that there is little variation
of the quantum defects, unlike what is observed for interacting bound series.
The calculated values are from the same model used to produce Fig.
21.15(b) (from ref. 1). show only a small variation as the 6p3/2l2s1/2 state
iscrossed. In a bound series this would produce one vertical cycle of the
plot, as in Fig. 21.2, not the curve of Fig. 21.16.l As isshown by the excellent
agreement between experiment and theory in Fig. 21.16, QDT predicts this
phenomenon. As pointed out by Cooke and Cromer,1 the lack of energy
displacements can be understood in a simple way. A 6p1/2ns1/2 state has a
second order repulsion from the 6p3/212sy2 state, which is distributed in
energy. Thus the contributions from the parts of the 6P3/212SX/2 state above
and below the 6p1/2fts1/2 state are opposite in sign. The resulting
cancellation of the energy shifts leads to the muted dispersion curve of Fig.
21.16. A second interesting aspect of interacting series is shown in Fig.
21.17, a plot of the scaled or reduced widths, i.e. v3F, of the 6p1/2ns1/2
states.1 They show a pronounced variation across the 6p3/2l2s1/2 state.
Because of the coupling of the 6p1/2fts1/2 and 6p3/212s1/2 states there are
two autoionization amplitudes coupling the nominal 6p1/2ns1/2 states to a
continuum. Schematically, these are 6p1/2ns1/2 �> / = 1 continuum, (21.23)
6p1/2ns1/2 �� 6p3/2�s1/2��/ = 1 continuum. ' Since the couplings are
to the same continuum, they interfere. As a result of this,if there is only one
continuum, at some point it is possible to have complete

448 Rydberg atoms 0.4 CO x Q LJJ O 0.2 Q LJJ DC QUANTUM NUMBER

Fig. 21.17 Observed (X) and calculated (�) reduced, or scaled, widths of
the Ba 6pmns1/2 states degenerate with 6p3/2l2s1/2 state. Note the
enhancement at n = 21, which originates in the constructive interference of
the two autoionization paths (from Ref. 1). destructive interference, resulting
in a vanishing autoionization rate. This phenomenon, first proposed by Cooke
and Cromer,1 is referred to as inhibited autoionization,18 stabilization,19
and bound states in the continuum?0 In zero field it is unlikely that the
energies will be such that complete cancellation of the autoionization rate
will occur. Nonetheless, van Woerkom etal.21 have observed a 190ns
lifetime of the Ba 5d3/2 26d5/2 / = 0state, which represents a decrease of
five orders of magnitude in the autoionization rate relative to that expected
for an unperturbed 5d3/2 nd5/2 J = 0series. Comparisons between
experimental spectra and R matrix calculations In this chapter we have
shown that complicated spectra can be understood using QDT. We have
considered examples of fitting spectra to QDT. Now we would like to
compare the results of Rmatrix calculations of spectra to the experimental
spectra in cases in whichfittingwould have been hopeless or nearly so.
Thefirstspectra which were synthesized using the Rmatrix approach were,
not surprisingly, vuv spectra, and synthetic spectra of Mg, Ca, Sr, and Ba
have all been calculated.22"25 In Fig. 19.2 we show the Ba vuv spectrum.
While the region above the Ba+ 6p1/2 limit is understandable, the region
below the limit exhibits
Comparisons between experimental spectra and R matrix calculations 449 o
35500 35700 35900 35500 35700 35900 35500 35700 35900 35500 35700
35900 Photon Energy (cm"1) Photon Energy (cm"1) Fig. 21.18 Effect of
overlap interference in the Mg 3pnd J = 3 spectrum: (a) synthesized spectrum
including interference, (b) same as (a) but using only direct-excitation terms,
(c) measured spectrum, and (d) corresponding overlap integral squared
(from ref. 12). clear evidence of interseries interaction, but defied further
analysis until the R matrix calculation of Aymar was able to reproduce it
almost perfectly.25 The R matrix method has also been applied with great
success to ICE spectra.12'26"29 An impressive example occurs in the
spectra of the Mg 3pnd J = 3 states.12'27 In Mg, because the Mg+ 3p1/2 and
3p3/2 limits are so close together (92 cm"1) the overlap integrals to the
channels converging to these two limits overlap substantially for n= 10 to n =
20 Rydberg states. As a consequence, nonzero excitations to channels
converging to both limits are the rule, and the spectra are almost impossible
to fit using approaches which work well for Sr and Ba. However, the R
matrix method worksvery well, even in this awkward case. In Fig. 21.18 we
show the spectrum from the Mg 3sl2d X D2 state to the Mg 3pnd J = 3
states.12 The experimental spectrum of Fig. 21.18 is obtained by exciting Mg
atoms in a beam to the 3sl2d X D2 state with two lasers, and scanning the
wavelength of a third laser across the region of the ionic 3s1/2 �� 3py
transitions while detecting the ions which result from the excitation of the
autoionizing 3pnd states. With all three lasers circularly polarized in the
same sense only J = 3 final states are produced. Inspecting Fig. 21.18, it is
not obvious how to describe the experimental spectrum in terms of states
converging to the 3p1/2 and 3p3/2 limits. Part of the problem derives from
the fact that these states are not particularly good jj coupled states, and part
from the fact that the overlap integrals to both ionic channels overlap, as
shown in Fig. 21.18. It is not a coincidence that these problems occur
together. The theoretical spectra are computed using a R matrix

450 Rydberg atoms 6p3/2 20s Fig. 21.19 The spectrum Ba 6p3/220s1/2 / = 1
state observed by ICE: (a) cross section o; (b)-(d) Anisotropy parameters /3
for electron ejection to the continua above the 6s, 5d, and 6p1/2 limits, (b)
6s1/2 limit, /?s; (c) 5d; limits, /?d; and (d) 6p1/2 limit, /?p; (e)-(g) Branching
ratios r for ejection of electrons to the continua above the (e) 6s1/2 limit, rs;
(f) 5d limits, rd; and (g) 6p1/2 limit, rp. The vertical line is drawn through
the maximum in the photoionization cross section, (from ref. 29).

References 451 computed by Greene,27 both with and without the

interference in the excitation amplitudes taken into account. (The K matrix
used in ref. 12 is identical to the R matrix defined in Chapter 20. If the
spectrum is computed without taking into account the interference in the
excitation amplitudes, aspurious feature at 35,700 cm"1 appears, which is
conspicuous by its absence in the experimental spectrum. When the
interference in the excitation amplitudes isincluded the calculated and
experimental spectra are nearly identical. In fact, Fig. 21.18istypical of
howwell the R matrix method reproduces the ICE spectra of Mg and shows
how the R matrix method allows the interpretation of spectra which would be
hard to unscramble otherwise. The Mg problem of Fig. 21.18 is not
intrinsically complex; there are only two continua. In Sr+ and Ba+, on the
other hand, there are lowlying dstates, and the analogous Sr 5pn� and Ba
6pn� states are degenerate with approximately ten continua. In spite of the
increased complexity, the R matrix method still performs well.28'29 An
illustration of how well is afforded by the comparison of Lange et
al.29between experimental observations of the ICE spectrum of the
Ba6sns1S0^> 6p3/22Os1/2 J=l transition and their counterparts calculated
by R matrix techniques. As shown by Fig. 21.19, not only does the total
photoionization cross section agree, but also the angular distributions and
branching ratios of the electrons to their possiblefinalstates. References 1.
W.E. Cooke and C.L. Cromer, Phys.Rev. A 32,2725 (1985). 2. U. Fano,
Phys.Rev. A 2, 353 (1970). 3. C.J. Dai, S.M. Jaffe, and T.F. Gallagher, /.
Opt. Soc. Am. B 6, 1486 (1989). 4. U. Fano, Phys.Rev. 124, 1866 (1961). 5.
S.A. Bhatti, C.L. Cromer, and W.E. Cooke, Phys. Rev.A 24,161(1981). 6.
N.H. Tran, P. Pillet, R. Kachru, and T.F. Gallagher, Phys.Rev.A 29,2640
(1984). 7. W.E. Cooke, S.A. Bhatti, and C.L. Cromer, Opt. Lett.7, 69 (1982).
8. S.A. Bhatti and W.E. Cooke, Phys.Rev. A 28,756 (1983). 9. N.H. Tran, R.
Kachru, and T.F. Gallagher, Phys. Rev. A 26, 3016 (1982). 10. J.G. Story and
W.E. Cooke, Phys.Rev. A 39, 4610(1989). 11. V. Lange, U. Eichmann, and W.
Sandner, /. Phys. B 22, L245 (1989). 12. C.J. Dai, G.W. Schinn, and T.F.
Gallagher, Phys. Rev. A 42, 223 (1990). 13. W. Sandner, G.A. Ruff, V.
Lange, and U. Eichmann, Phys.Rev.A 32, 3794 (1985). 14. R. Kachru, H. B.
van Linden van den Heuvell, and T.F. Gallagher, Phys.Rev.A 31, 700 (1985).
15. T.F. Gallagher, J. Opt.Soc. Am. B 4, 794(1987). 16. F. Gounand, T.F.
Gallagher, W. Sandner, K.A. Safinya, and R. Kachru, Phys.Rev.A 27, 1925
(1983). 17. J.P. Connerade, Proc. Roy. Soc. {London)362, 361 (1978). 18. J.
Neukammer, H. Rinneberg, G. Jonsson, W.E. Cooke, H. Hieronymus, A.
Konig, K. Vietzke, and H. Springer-Bolk, Phys.Rev. Lett.55, 1979 (1985).
19. S. Feneuille, S. Liberman, E. Luc-Koenig, J. Pinard, and A. Taleb,/.
Phys.B 15,1205 (1985). 20. H. Friedrich and D. Wintgen, Phys.Rev. A 32,
3231 (1985). 21. L.D. van Woerkom, J.G. Story, and W.E. Cooke, Phys.Rev.
A 34, 3457 (1986). 22. P.F. O'Mahony and C.H. Greene, Phys.Rev. A 31,250
(1985).

452 Rydberg atoms 23. C.H. Greene and L. Kim, Phys.Rev. A 36, 2706
(1987). 24. M. Aymar, /. Phys. B 20, 6507 (1987). 25. MAymar, /. Phys. B
23, 2697(1990). 26. V. Lange, U. Eichmann, and W. Sandner, J. Phys.B 22, L
245 (1989). 27. C.H. Greene, private communication. 28. M. Aymar and J.
M. Lecompte, /. Phys. B. 22, 223 (1989). 29. V. Lange, M. Aymar, U.
Eichmann, and W. Sandner, /. Phys. B. 24, 91 (1991).

22 Interseries interaction in bound states Perturbed Rydberg series One of the

most intensively studied manifestations of channel interaction in the bound
states is the perturbation of the regularity of the Rydberg series, which is
evident if one simply measures the energies of the atomic states. Although
measurements of Rydberg energy levels by classical absorption spectroscopy
showtheperturbations inthe serieswhich areoptically accessiblefrom the
ground state, the tunable laser has made it possible to study series which are
not connected to the ground state by electric dipole transitions as well. One
of the approaches which hasbeen usedwidelyis that used byArmstrong etal.1
As shown in Fig. 22.1, a heat pipe oven contains Ba vapor at a pressure of
~1 Torr. Three pulsed tunable dyelaserbeamspassthrough
theoven.Twoarefixedin frequency, to excite the Ba atoms from the ground 6s2
^Q state to the3Pt state and then to the 6s7s3Sx state. The third laser is
scanned in frequency over the 6s7s3SX �> 6snp transitions. The Ba atoms
excited to the 6snp states are ionized either by collisional ionization or by the
absorption of another photon. The ions produced migrate towards a
negatively biased electrode inside the heat pipe. The electrode has a space
charge cloud of electrons near it which limits the emission current. When an
ion drifts into the space charge region it locally neutralizes the space charge,
allowing many electrons to leave the region around the negative electrode,
and this current pulse is detected.2 In other words the device is a space
charge limited diode. The experiments have not been confined to the resonant
excitation scheme described above; off resonant multiphoton excitation has
been used extensively. Camus et al. have used an optogalvanic cell in which
a glow discharge is maintained in a mixture of He and Ba.3 In the discharge a
substantial population accumulates in the metastable Ba 6s5d 3 Dy levels,
and from these levels it is straightforward to reach the Rydberg levelswith
twopulsed dye laser photons. A Ba atom in aRydberg state is much more
easily ionized than aBa atom in alower lying state, and as a result, whenever
a Rydberg state is populated the current in the discharge increases
temporarily, and this increase is detected. Finally, thermal beams of atoms
can be excited using one or more lasers, as described inChapter 3. An
interestingvariant is themetastable beamusedbyPost etal.4A Babeam from an
effusive oven is used to conduct a current from aheated 453

454 Rydberg atoms FABRY-PEROT REFERENCE IONIZATION OVEN

PHOTODIODE ' i ' SIGNAL AIR-SPACED FABRY-PEROT IONIZATION
SIGNAL (b) S $ $ ^ ^ ^ ^ \ ^ IONIZATION LIMIT msnp 3 P� I/3 (TUNED)
ms(m+l)s 3 S, V2 (FIXED) msmp 3 P? Vy (FIXED) ms2 Fig. 22.1(a)
Experimental setup for three laser excitation of atomicvapors with ionization
detection andreference signalfor calibration, (b)Energy leveldiagram ofa
typicalalkaline earth atom showing the sequence which allows easy
excitation of 3 P�states (from ref. 1). filament several mm from the oven
orifice. A 20 V potential difference between the two produces a 400 mA
current through the atomic beam, mostly carried by electrons. The electrons
excite the Ba atoms, and some of the atoms decay to metastable states, which
remain excited some distance downstream from the oven. The metastable Ba
6s5d 3 D7 state is readily populated in this way. Downstream, the Ba atomic
beam is crossed at 90� by the beam from a frequency doubled, single mode,
cwdye laser, which excites the metastable Ba atoms to the 6snp and 6snf
Rydberg states. The Rydberg atoms pass out of the excitation region and into
the region between a pair of field plates, where they arefield ionized, and the
resulting ions detected. All of these approaches have energy resolution
limited by the laser Hnewidth, �0.3 cm"1 for the pulsed lasers and 2 MHz
for the cw laser. The intensities of the observed transitions cannot, however,
be used with complete confidence, for itis
Perturbed Rydberg series 455 ^ 0.25 o 1 1 , � ! a � 1 1 D a 3 D a 0.5 0.75
1.0 0 0.25 0.5 0.75 1.0 Fig. 22.2 Lu-Fano plot of the Ba / = 4 levels below
the 5d3/2 limit: 5d3/2rcd3/2 (�), 5d5/2>zd5/2 (D), 5d5/2rcd3/2 (�) (from
ref 5). not clear that the efficiency of detection of all Rydberg states is the
same. To the extent that collisional ionization isimportant, the radiative
lifetimes of the states are significant, for longer lifetimes are more likelyto
allowionization. Aswe shall see these vary dramatically in the vicinity of
perturbations in the spectra. If photoionization of the Rydberg states plays a
role, the cross section depends on both the wavelength and the amount of
Rydberg and perturber character in the atomic states under study. The
measured energy levels cannot be expressed as with a constant quantum
defect. Instead, analysis of the data requires two steps. First it is important to
decide how many channels are involved and how many ionization limits.
Once the identity of the important limits is established each observed energy
can be assigned an effective quantum number relative to each limit.For
exampleifthere aretworelevant limits, 1 and2, eachobserved energyis
assigned the quantum numbers v1 and v2, and these canbeplotted as shownin
Fig. 22.2, a plot by Aymar etal.5 of the effective quantum numbers
corresponding to the observed term energies of the 5d7/id;/ = 4levels below
the 5d limits of Ba+.6 Although allthe levels studied are above the Ba+
6s1/2limit, autoionization to the

Rydberg atoms 4.75 Fig. 22.3Lu-Fano plot of the high lying Ba even parity
7=2 levels above the 6s22d3D2 level. The full curve iscalculated with the
QDTparameters of ref. 8: observed6snd3D^ states (+), observed 6snd1D2
states (�); observed 5d7dperturber (V) (from ref. 8). 6ss�continua plays no
role in the spectra at aresolution of0.3 cm"1. Plots such as Fig. 22.2 are
called Lu-Fano plots.7 The points represent the intersection of the quantum
defect surface and the energy constraint. Using the observed points itis
possible tofindthe parameters which generate the quantum defect surface
which passes through the points. A useful observation, of Lu and Fano, is that
the diagonal line from vt = 0, v2 =0 to vt = 1, v2 = 1 intersects the quantum
defect surface at the values of jua, i.e. vx = v2 = jua-Fig- 22.2 is an example
of a three channel, two limit problem. If there are more limits the point must
be plotted in more dimensions, although if there are only more channels, but
still two limits a twodimensional graph canstill
beused;itsimplyhasmorebranches.Fig.22.2isin several ways an excellent
example of a Lu-Fano plot. It is plain to see that the 5d3/2ftd5/2 series has a
quantum defect of 0.65 and that the 5d5/2nd/ series have quantum defects of
0.60 and 0.68 where they are not interacting and that the interaction between
the series causes local deviations from these values. It is not alwaysthe
casethat the Lu-Fano plot consists of horizontal and vertical lines with
localized avoided crossings. In their study of the 1 P1 odd states of alkaline
earth atoms Armstrong et al.l obtained Lu-Fano plots with no horizontal or
vertical sectionsindicating that the entire series oflVx states converging to the
lowestlimit contained significant admixtures of doubly excited states. The
relatively well localized interaction of the Ba 5d7d X D2state with the Ba
6snd 1 '3D2states is shown by the section of the Lu-Fano plot of Fig. 22.3.8
The

Properties ofperturbed states 457 5d7dperturber is onlymixed significantly

into a fewstates, and onlyone branch of the Lu-Fano plot is significantly
perturbed. However, as shown by Fig. 22.3,the branch which has v6s ~ 0.3 at
energies below the 5d7d perturber evolves into the branch with v6s ~ 0.2
above the perturber, and as we shall see, the X D2 and3D2 states interchange
character at the 5d7d X D2state. Properties of perturbed states The energy
level variations exhibited in the Lu-Fano plots are reflected in analogous
variations in the atomic properties. One of the most apparent variations isin
the lifetimes of Rydberg states in the vicinity of aperturber. Since the
perturbing states tend to be relatively compact doubly excited states, they
have larger matrix elements for short wavelength transitions to lower lying
states than do the Rydberg states, and their lifetimes are reduced accordingly.
Although the radiative lifetimes in the regions around several perturbing
states have been measured,9"12 we here focus on the variation in the
lifetimes near the Ba 5d7d state perturbing the 6snd1'3D2series. The
lifetimes have been measured bytwotechniques. Thefirst,used by Aymar
etal.10 and Gallagher et al.11 is to use two pulsed lasers to excite Ba atoms
in a thermal atomic beam to the 6snd Rydberg states via the route
6s21S0��6s6p 1 P1 �> 6snd13D2. The population in the Rydberg state as
afunction of time after excitation is determined byapplying afieldionization
pulse at avariable time after the laser pulse. The second method, used by
Bhatia et al.12 isto use two cwlasers to excite Ba atoms by the same route.
The second 6s6p 1 P1-6snd 13 D2 laser is pulse modulated, at a high
repetition rate, and the temporal decay of the 6srad-6s6pfluorescenceis
recorded after thebluelaseris pulsed off. Both methods givesimilar results.
For example,the lifetimes reported inrefs. 10-12for the5d7d 1 T>2 state are
160,217, and 170 ns,respectively. If weplot the decay rates, instead of the
lifetimes of the states, we can see in a very direct way the amount of 5d7d X
D2 character which exists in each state, and in Fig. 22.4 we show a plot of
the decay rates as a function of term energy.11 As shown by Fig. 22.4,
roughly five states have decay rates significantly higher than the lines
representing the expected n~~3 variation in the decay rates of the 6snd1'3D2
Rydberg states. Note that it is the singlet states on the low energy side of the
perturber which have the increased decay rates and the triplet states on the
high energy side, as expected from the Lu-Fano plot of Fig. 22.3. Although
Aymar etal.10 have made a more sophisticated analysis of the lifetime
variation, a straightforward method is to describe the ith state near the 5d7d
state by the wavefunction11 % = ^ 5 d 7 d + (1 - �01/2^6snd , (22.2) and the
radiative decay rate by

458 Rydberg atoms 41880 41840 41800 TERM ENERGY (cm"1) Fig.
22.4Observed decay ratesof the 6snd *D2 (O), 6sml3D2 (�), and5d7dlT>2
(A) states. Thelinesindicate the expected v^ dependence for the unperturbed
6srcd*D2 and6srcd3D2 Rydberg series. For points without error bars shown
they are approximately the size of the symbols (from ref. 11). F = e F +
(1��-)F (22 3) If we approximate T6snd byits average, 0.35 x 106s"1,
over the 24 < n < 29 range and use the fact that " i = U (22.4) i.e. that there is
one5d7dstate spread among all thestatesshownin Fig.22.4,then wecan easily
solve Eqs. (22.3) and (22.4) for the decay rate of the pure 5d7d state and the
fraction e{ of 5d7d state in the perturbed 6sndRydberg states. The pure 5d7d
decay rates of 1.5 X 107 s"1 and 1.2 x 107 s"1 were obtained by Aymar et
al.10 and Gallagher et al.n More interesting are the values of eb the
fractional perturber character in each state. In Table 22.1 we give the
perturber fractions obtained from the lifetimes as well as those obtained by a
three channel QDT analysis of the energies of the states. The lifetime and
QDTresults are certainlyin reasonable agreement, and the experimental
results can be read approximately from the graph of Fig. 22.4 using et = 8.3 x
1(T8 s(F - 0.45 x 106s"1). (22.5) In the Ba 6snd 1 '3D2series, the higher
energy levels are labelled singlets, as shown by Figs. 22.3 and 22.4.
According to the QDT analysis,8 the states on the continuous branch of the
Lu-Fano plot of Fig.22.3 evolvefrom singletsbelow the 5d7d state to triplets
above the 5d7d state and contain negligible 5d7d character. The states on the
broken branch change from predominantly triplet character to singlet
character at the 5d7d state. This assignment of singlet and triplet character is
borne out to some extent by the line strengths in absorption spectra.13 It is

Properties of perturbed states 459 1.20 2.60 Fig. 22.5 The g factor of Sr 5snd
states as afunction of vD3/2, the effective principle quantum number
measured relative to the 4d2D3/2 ionization threshold at 60488.09 cm"1. The
solid lines are the theoretical predictions, and the points correspond to the
experimental measurements for the bound states designated by 5snd (from
ref. 15). Table 22.1. Perturber fractions obtained from the radiative decay
rates, photoionization cross sections, and QDT analysis of the optical
spectrum. State 6s25d 3 D2 6s26d 3 D2 5d7d *D2 6s27d X D2 6s28d *D2
6s29d *D2 Perturber fraction e derived from Experimental radiative decay
rates'* 0.035 0.219 0.365 0.262 0.068 0.028 Experimental photoionization
cross sections^ 0.04(2) 0.11(4) 0.45(6) 0.25(5) 0.07(3) 0.05(2) QDT three
channel model 6s1/2, 5d5/2 limits" 0.02 0.13 0.40 0.20 0.05 0.02 a (from
ref. 11) b (from ref. 16) demonstrated explicitly, however, in an analogous
case in Sr.1415 From an analysis of the energy levels of the Sr 5snd 1 3 D2
states Esherick predicted that the singlet and triplet character of the Sr 5snd 1
3 D2 states interchanged at n = 15.14 Wynne et aL15 then verified this
prediction experimentally by measuring the g factors of these states. The
singlet and triplet g factors are 1and 7/6 respectively,

460 Rydberg atoms so an accurate measurement of the g factor gives a more

reliable indication of singlet and triplet composition than do the line
strengths, for the composition of the initial state plays no role in the g factor
measurements. The experiment was done byplacing a heat pipe, such as the
one showninFig. 22.1, between the poles of an electromagnet. A single
linearly polarized laser was used to drive two photon transitions from the Sr
5s21S0 ground state to the 5sndstates. With the laser polarization
perpendicular to the magnetic field, each 5s2-5s/id resonance was split into
three resonances, corresponding to the allowed m = 0, �2 final states.
Measuring the relative splittings of the levels in a magnetic field of 8.3 kG
yielded the relative g factors, and assigning the 5sl2d *D2state its calculated
g factor of 1.0033 leads to the absolute values shown in Fig. 22.5.15
Measuring the field independently could onlybe done to 3%, which gave an
absolutevalue of the 5sl2d l T>2 g factor of 1.035(37), in agreement with the
calculated value. In any event, it is clear from Fig. 22.5 that the singlet and
triplet characters evolve as shown bythe solid lines of Fig. 22.5, which are
predictions from the QDT analysis of the energies only.14 In Ba, many of the
perturbers of the bound Rydberg series are low lying states converging to the
5dy limits of Ba+ and are well described as5dn� states oflow n.
Accordingly, it is reasonable to expect that they should have large cross
sections for the transition to the autoionizing 6pra� state which may
expected to be rather broad sincenis low.The 5dn��> 6pn�transitions are
found inthe general vicinity of the Ba+ 5d�� 6ptransitions, at X ~ 630 nm.
In contrast to the perturbing 5dn( states, the Ba 6sn( Rydberg states are
unlikely to absorb 630 nm light. By measuring the relative photoionization
cross sections at X ~ 630 nm Mullins et al.16 measured the perturber
character of bound Ba 6sns and 6snd Rydberg series near several 5dnt
perturbers. Their resultsfor theBa6snd1'3D2states aregiven in Table 22.1.
Continuity of photoexcitation across ionization limits One of the aspects of
QDT which makes it most useful is that it clearly shows the continuity of
many properties across ionization limits. For example, if there are two
coupled Rydberg series converging to two different ionization limits, the
effective quantum numbers are vx and v2 below the first limit. Between the
two limits the spectrum is composed of a series of Beutler-Fano profiles
corresponding to the Rydberg series converging to the higher limit, and
plotted on a scaleof v2 instead of energy, the spectrum repeats modulo 1 in
v2. According to QDT the spectrum also repeats modulo 1 in v2 below the
first limit. More precisely, since the spectrum belowthefirstlimitis composed
of discretelines,the envelope of the spectrum repeats. An excellent
illustration ofthis notion occursinthe spectrum of

Forced autoionization 461 the Ba 5d/zdy/ = 4 states asthe Ba+ 5d3/2limit is

traversed.17 These are the states shown in the Lu-Fano plot of Fig. 22.2. In
Fig. 22.6(a) we show the spectrum of the transition from the Ba 5d6pXF3
state to the 5d5/216d3/2?5/2stateswhich liejust abovethe Ba+5d3/2 limit,
andin Fig. 22.6(b) we show the spectrum from the same initial state to the
5d5/2l4d3/2?5/2 states which liejust below the Ba+ 5d3/2 limit. It
isapparent that the envelopesof the excitations are identical. We have been
discussing these states asif they were bound states, and in fact for our present
purpose they might as well be. First, below the 5d3/2 limit the observed
linewidths equal the laser linewidth, and second, there is no visible
excitation of the 6se� continua belowthe5d3/2limit. It is also useful to note
that a three channel quantum defect treatment of the 5dynd; 7=4 series
reproduces both the Lu-Fano plot of Fig. 22.2,and the spectra shown in Fig.
22.6. The experimental spectra were reproduced by QDT models using both
the a and / channel dipole moment parametrizations described in Chapter 21.
Forced autoionization Above the ionization limit interchannel interactions
lead to autoionization and below the limit they lead to series perturbations.
Forced autoionization allows both manifestations to be observed in the same
states. The essential idea is shown inFig.22.7.In zerofieldastate converging
to a higher limit liesjust belowthe first limit and perturbs the Rydberg series
converging to thefirstlimit. If we apply an adequate
staticelectricfieldwedepress the lower ionization belowthe perturbing state
so that it interacts with a Stark induced continuum of states. In this case the
perturbing level appears as an autoionizing resonance. This phenomenon was
termed forced autoionization by Garton et al.ls who first observed it in the
spectrum of shock heated Ba. Since that time it hasbeen studied in a
quantitative fashion with both static19"22 and microwave23 fields. In their
systematic experiments Sandner etal20'21 used two tunable dye lasers to
excite ground state atomsin aBa beam byeither of twooptical paths, A and B,
to the vicinity of the 5d7d X D2perturber, as shown in Fig. 22.8. In the
absence of an electric field the Ba atoms excited to the 5d7d or neighboring
6snd Rydberg stateswere ionized byafieldionization pulse and the
ionsdetected. If astatic field was present it alone ionized the atoms. In Fig.
22.9 we showthe excitation spectra using both paths A and B with and
without a static field. Consider first the zero field spectra. Usingpath A,
viathe 6s6p1F1 state, the 5d7dstate is not excited and we see weaker
transitions to states near 41841 cm"1 which have appreciable 5d7d
character. In contrast, using path Bvia the 5d6p3DX state, only those states
with appreciable 5d7d X D2character are excited. In the E # 0spectra of Fig.
22.9 it is apparent that for path A we have a q ~ 0Beutler-Fano profile, and
for path Ba

462 Rydberg atoms 1.0 CO CO (a) I 0.0 47080 47090 47100 47110 47120
Energy (cm 1 ) 1.0 o.o 46820 46840 46860 46880 46900 Energy (cm"1) Fig.
22.6 (a) Observed Ba 5d6pl F3-5d5/216d3/25/2 spectrum. The strong
doublet of broad / = 4 lines is located at 47093 and 47100 cm"1. Two weak /
= 3features are visible at 47102 and 47103 cm"1, (b) The transition from
5d6p ^3 to the 5dnd 7 = 4 levels below the 5d3/2 limit showing the same
basic oscillator strength distribution to the 5d5/214d3/25/2 states over many
essentially discrete energy levels (from ref. 17).

Forced autoionization 463 E=0 E*0 Fig. 22.7 Schematic diagram of forced
autoionization. Above the ionization limit the autoionizing state A,
converging to a higher limit, is manifested as an autoionization resonance.
Belowthelimit, in zerofieldtheinteraction oftheperturber Pwiththe Rydberg
series results in the perturbation of the Rydberg series. Applying an electric
field E depressestheionization limitbelowP, anditappears as a forced
autoionization resonance. 6s 6s Fig. 22.8 The two different excitation
schemes A and B used to observe the Ba 5d7d perturber as a forced
autoionization resonance. Path A leads to a q parameter near zero, while B
leads to a large q parameter (from ref. 21). profile with large \q\. In both
cases the field has converted the 5d7d state into an autoionizing resonance,
as expected. However, even a casual inspection of Fig. 22.9 reveals that the
path B zero field spectrum is not simply a discretized version of the E = 7.5
kV/cm spectrum; it is broader. The broadening is in fact an experimental
artifact introduced by the 500 ns delay in applying the field ionization
pulse.21 When the short, �200 ns, lifetimes of states with appreciable 5d7d
character are taken into account to extrapolate the zero field spectrum of Fig.
22.9 back to zero time delay, the intensities of the central lines are increased

464 Rydberg atoms c D _O D in c Q) -4�' c Jilll 1 L.lili Ik i i (a) OV/cm 5 d

7 d 1 D 2 1 �-o (b) 4.8kV/cm -0 0.. (c) OV/cm (d) 7.5kV/cm �1600
�2000 �2400 �2000 �2 �00 E(cm"]) Etcnrf1) Fig. 22.9 (a)
Photoabsorption spectrum observed in zerofield,using excitation path Aof
Fig.22.8. (b) Same as(a),butobserved at anapplied dc electricfieldof4.8
kV/cm. (c)Zero field spectrum observed usingpath B of Fig. 22.8. (d) Same
as (c),but with an applied field of7.5kV/cm(fromref. 21). by a factor greater
than 10, and the zero field path B spectrum appears to be a discretized
version of the E = 7.5 kV/cm spectrum.21 We can compare the forced
autoionization resonance to the predictions of zero field QDT. The observed
width, 15.5 cm"1, and the width from QDT, 15.3 cm"1, are in excellent
agreement. However, the energy positions are different by 5 cm"1. Exactly
whyisnot clear, but it iscertainly the case that the Stark induced continuum is
not in all respects like a zero field continuum. For example, with both lasers
polarized parallel to the field, so as to excite m =0 final states, the forced
autoionization resonance analogous to the one shown in Fig. 22.9 exhibits
structure due to the long lived, blue shifted, Stark states.20'21 References 1.
J. A. Armstrong, J. J. Wynne, and P. Esherick, /. Opt.Soc. Am. 69, 211
(1979). 2. K. H. Kingdon, Phys.Rev. 21, 408 (1923). 3. P. Camus, M.
Dieulin, and A. El Himdy, Phys. Rev. A 26, 379 (1982). 4. B. H. Post, W.
Vassen, W. Hogervorst, M. Aymar, and O. Robaux, /. Phys.B 18, 187 (1985).
5. M. Aymar, P. Camus, and A. El Himdy, Physica Scripta 27,183(1983). 6.
P. Camus, M. Dieulin, A. El Himdy, and M. Aymar, Physica Scripta 27, 125
(1983). 7. K. T. Lu and U. Fano, Phys.Rev. A 2, 81 (1970).

References 465 8. M. Aymar and O. Robaux, /. Phys.B 12,531 (1979). 9. M.

Aymar ,P. Grafstrom, C. Levison, H. Lundberg and S. Svanberg, /. Phys. B
15, 877 (1982). 10. M. Aymar, R. J. Champeau, C. Delsart, and J.-C. Keller,
/. Phys.B. 14,4489(1981). 11. T. F. Gallagher, W. Sandner, and K. A. Safinya,
Phys.Rev. A 23, 2969 (1981). 12. K. Bhatia, P. Grafstrom, C. Levison, H.
Lundberg, L. Nelsson, and S. Svanberg, Z. Phys. A 303, 1 (1981). 13. J. R.
Rubbmark, S. A. Borgstrom, and K. Bockasten, /. Phys.B 10,421 (1977). 14.
P. Esherick, Phys.Rev. A 15, 1920 (1977). 15. J. J. Wynne, J. A. Armstrong,
and P. Esherick, Phys.Rev. Lett.39, 1520 (1985). 16. O. C. Mullins, Y. Zhu,
and T. F. Gallagher, Phys.Rev. A 32, 243 (1985). 17. C. J. Dai, S. M. Jaffe,
and T. F. Gallagher, /. Opt.Soc. Am. B 6,1486(1989). 18. W. R. S. Garton, W.
H. Parkinson, and E. M. Reeves, Proc. Phys.Soc.80, 860 (1962). 19. B. E.
Cole, J. W. Cooper, and E. B. Salomon, Phys.Rev. Lett. 45, 887 (1980). 20.
W. Sandner, K. A. Safinya, and T. F. Gallagher, Phys.Rev. A 24, 1647
(1981). 21. W. Sandner, K. A. Safinya, and T. F. Gallagher, Phys.Rev.A 33,
1008 (1986). 22. T. F. Gallagher, F. Gounand, R. Kachru, N. H. Tran, and P.
Pillet, Phys.Rev. A 27, 2485 (1983). 23. R. R. Jones and T. F. Gallagher,
Phys.Rev. A 39,4583 (1989).

23 Double Rydberg states The autoionizing two electron states we have

considered so far are those which can be represented sensibly by an
independent electron picture. For example, an autoionizing Ba 6pnd state is
predominantly 6pndwith only small admixtures of other states, andthe
departures from theindependent electron picture canusually be described
using perturbation theory or with a small number of interacting channels. In
allthese cases one ofthe electrons spendsmost ofitstimefar from the core, in a
coulomb potential, and the deviation of the potential from a coulomb
potential occurs only within a small zone around the origin. In contrast, in
highly correlated states the noncoulomb potential seen by the outer electron
is not confined to a small region. In most of its orbit the electron does not
experience a coulomb potential, and an independent electron description
based on n�n'� r states becomes nearly useless. There are two ways in
which this situation can arise. The first, and most obvious, is that the inner
electron's wavefunction becomesnearly aslarge as that oftheouter
electron.Ifwe assign the twoelectrons the quantum numbers n{�{and no(o,
thisrequirement is met when nx approaches no, which leadstowhat might be
called radial correlation. The sizes of the two electron's orbits are related.
The second way the potential seen by the outer electron can have a long
range noncoulomb part is if the presence of the outer electron polarizes the
inner electron states. This polarization occurs most easily if the inner
electron �states of the same n are degenerate, for then even a very weak
field from the outer electron, equal to 1/r^, converts them to Stark
states,whichhavepermanent dipolemoments.The anisotropicpotential from the
induced dipoleproduces an angular correlation in themotion of thetwo
electrons, even if the orbit of the outer electron is many times larger than the
orbit of the inner electron. Angular correlation may be found in all the doubly
excited states of He since He+ is hydrogenic. On the other hand, radial
correlation is presumablyonlyfound when nx approaches no. Angular
correlation may also be observed in the doubly excited states converging to
the nearly degenerate high �states of a singly charged ion, Sr+ for example,
even when the outer electron is in a substantially larger orbit than the inner
electron. On the other hand, the doubly excited statesconverging
tolow�statesof Sr+, whichhavelargequantum defects, do not exhibit angular
correlation before they exhibit radial correlation. To see remarkable
correlation in these states, the requirement that nx approach no must be met.
466

Doubly excited states of He Doubly excited states of He Although the

existence of autoionizing states has been known since the 1920s,it was the
observation of the doubly excited states of He by Madden and Codling1
which aroused interest in highly correlated states. Using synchrotron
radiation, they recorded the absorption spectrum of He using an 83 cmpath
length ofHe at a pressure of 0.3 Torr. One might naively expect to see two
series, 2snpand 2pras, both converging to the He+ n = 2 limit. However, they
observed one strong series, the + series, and a very weak series, the -
series.2 In addition to being weaker, the � seriesliesbelowthe +
seriesandhassmaller autoionization widths. More recently the measurements
have been repeated and extended by Domke etal.3 who also used a
synchrotron as a light source, but detected the photoions produced instead of
the absorption of the synchrotron radiation. Their spectra are shown in Fig.
23.1. As shown by Fig. 23.1 the strong n =2 + peaks appear as classic
Beutler-Fano profiles on top of a small background lsep photoionization
signal. They are roughly 100 times as intense as the � resonances. In
addition, the widths of the � peaks are about 10% of the widths of the +
peaks. If we examine the spectra converging to the higher He+ limits, shown
in Fig. 23.2, several features are apparent. First, wecan seethat these spectra
contain only oneseries, the + series. In addition, inthe spectra converging to
higher nstates of the ion the resonances become smaller relative to the
background photoionization, which is hardly surprising, as there are more
available continua. Stated another way, the magnitude of the Fano q
parameter decreases with increasing n, although it also changes sign, which
isnot as simply explained. In Fig. 23.2(c) wecan see that the
seriesconvergingtothe He+ n = 5 and 6 andn = 6 and7 statesoverlap,
producing the same sort of pattern seen in Fig. 19.2. The initial experiment of
Madden and Codling, the observation of the states converging to the He+ n =
2 state, was first explained by Cooper et al.2 in the following way. Rather
than use the 2srap and 2prcs states as the atomic basis, they proposed using
the combinations2 1 ' �u2pm] (23.1) where u(2snp) is a product of a 2s and
an np wavefunction. The + and - states have characteristics which are
different from those of angular momentum states. In the + statesthe radial
oscillationsof thetwo electrons areinphase, andin the states they are out of
phase. In addition, by considering the properties of the + and � states near
the origin we can explain the n =2spectra, shown in Fig. 23.1. Both u2snp
and u2pns are products of one s wavefunction and one p wavefunction. Thus
at the origin they are functionally the same. Since the normalization factors
ofwavefunctions ofprincipal quantum number 2 andnaregivenby2~3/2and
n~m respectively, the overall normalization factors of u2snp and u2pns are
nearly the same. Near the origin u2snp ~ u2pns, and W_ nearly vanishes. As
a result, the states are only weakly excited from the ground state, and they
have lower autoionization rates than the + states, as observed.

468 Rydberg atoms i�[//

�i�i�i�i�i//i�i�i�i�i�I//T�i�i�i�iJ�i�i�i�i�I//J�i
i�i�i i// i i i i y/j i i i i \/j\ i i 62.75 62.78 64.13 64.16 64.64 64.67 64.91
64.94 65.04 65.07 Photon Energy (eV) Fig. 23.1 Autoionizing states of
doubly excited Hebelow the He+ n = 2 threshold: (a) overview, (b)
magnification of the n > 6region, and (c) In- states (from ref. 3).

Theoretical descriptions 469 69 7 0 7 1 7 2 7 3 7 2 . 4 7 2 . 6 7 2 . 8 7 3 . 0 7

3.574.074.575.075.575.375.475.575.576.076.57
7 . 0 7 7 . 5 Photon Energy (eV) Fig. 23.2 Autoionizing states ofHe: (a)
below the He+ n= 3threshold, (b) below the He+ n = 4 threshold, and (c)
below the He+ n= 5 and6 thresholds. The high n regions are shown magnified
on the right-hand sides in (a) and (b). Note the overlapping ofseries in (c)
(fromref. 3). Theoretical descriptions The work of Cooper et al.2 was the
beginning of a steady stream of theoretical work. Shortly after the
experiments of Madden and Codling more detailed

470 Rydberg atoms ie2 He2+ Fig.23.3 TheHe atom. calculations of the
doubly excited He levels were carried out by a variety of techniques.4"7
While these calculations showed that the description of Cooper et al?
omitted aspects of the problem, such as the contribution of the 2pnd states,7
the basic feature of their explanation, the creation of superposition states,
somewhat similar to Stark states, was found in all these treatments.4"7 While
the detailed calculations gave energies and widths in reasonable agreement
with the observed experimental results, they did not provide a simple
physical picture of why the He spectrum was the way it was. To address this
issue Macek described theHe atom usinghyperspherical coordinates. Inthis
approach, the two electrons, each described by three coordinates, are
replaced by an equivalent singleparticle insix dimensions.8'9InFig. 23.3 we
showthe He atomin which rx and r2 are the vectors from He2+ to each of the
two electrons. The hyperradius Rh is defined by Rh = vTfTTL (23.2) and the
ratio of rx to r2 defines the hyperangle a, i.e. tana = r2/r1. (23.3) At a = 0and
nil electrons 1 and 2, respectively, are infinitely far from the He2+, while at
a =JIIA they are equidistant from the He2+. Following the approach of
Macek8 and Fano,9we replace the usual wavefunction xp by ^, which is
related to ^(r,a,r1?f2) = sin a cos aR^2xp{R^a^r2), (23.4) where the unit
vector r{ = rjr^ This substitution is roughly analogous to replacing the radial
function R(r) by p(r)lr in the H atom. Using W the Schroedinger equation in
hyperspherical coordinates is given by9 a,r1?f2) - E ^(flh,a,f 1?f2) = 0 (23.5)
where A is the grand angular momentum of Smith10 defined by A2= ^=- - - +
�V- + -^-' (23-6) da 4 cos a sin a

Theoretical descriptions 471 cos0. cos0a y \y/ 75' 60' 45' 30* Fig. 23.4
Three dimensional plot of -C(a,012) with Z = 1 in hyperspherical
coordinates; the ordinates represent apotential surface inRydberg units at Rh
= la0 (from ref. 11). where �x and �2 are the angular momenta of the two
electrons. From Eqs. (23.5) and (23.6) it is apparent that A2 plays the role of
�2 in the H atom. A key point of the hyperspherical approach is that the
potential V can be factored; V(Rh9a9tl9t2)=^-C(a9tl9t2)9 (23.7) where The
scaled potential C(a,f1?f2) is shown in Fig. 23.4 using cos 012 = h'h-11
Inspecting it we can see that C is flat at a = 45�, corresponding to rt = r2,
for all angles 012 except 012 = 0. The potential valleys, at a = 0 and nil
correspond to one of the two electrons being far from the He2+ core and the
other being close to the core. Inspecting Fig. 23.4we can see that there is
asaddle point in C(a,r1 ,f2) at 012 = Jt and a = JT/4. If we define the
operator ! ^ ( � A A ) (23-9) and solve the eigenvalue equation a9tl9t2) =
UAI(/?h)0AI(a,f1,t2) (23.10) at constant Rh, we find the eigenvalue t/^(i?h)
corresponding to the solution (pM with the set of quantum numbers//
corresponding to the a, f1? and f2 coordinates.

472 Rydberg atoms An alternative approach is to use the well known

eigenfunctions of A2 and cope with off diagonal matrix elements due to
C(a,f1?r2). If we now assume that a,tl9t2) = Y WWaAA), (23.11) JU where
F^(/?h) is a wavefunction analogous to the vibrational wavefunction in a
diatomic molecule, Eq. (23.5) becomes [ ^ ^ 2 +HR]F�( R h)<t>� R
WFM(Rh)</>r(a9tl9t2). (23.12) If we replace HR(pM by VM(Rh)<j)M, use
the orthogonality of the 0^ functions, and make the adiabatic assumption that
d(pjU(a,rur2)/dRh is negligible, wecan remove the angular variation from
Eq. (23.12), leaving the radial equation % ~w]F� (Rh) = � (23 -13) in
which L^(i?h) is the effective radial potential, which plays the same role as
the internuclear potential curve in a diatomic molecule. In fact, the adiabatic
approximation is in essence the same approximation as the Born-
Oppenheimer approximation used to treat diatomic molecules. In the
descriptions of a diatomic molecule the fact that the nucleii are more massive
is usually given as the reason why the separation works. However, all that
isreally necessary is that the motion in the a,f1?r2 coordinates not change
rapidly with Rh. In Fig. 23.5 we show the set of potential curves calculated
by Sadegphour and Greene for H~, which differs from He in having
shallower wells.12 Each curve corresponds to a set ofju quantum numbers.
InFig. 23.5(a) we canseethat thereis afamily of curveswhich connectstoH
levelsof different n atRh �> o�. We can also see that as Rh is reduced from
infinity that the potential curves from each n split into a fan of curves,
withlarger spacingathigher n. This splittingis a manifestation of the Stark
effect from thefieldof the well removed electron on the remaining H atom. It
is also apparent that only the lowest lying curves have wells at small Rh
allowing the formation of quasi-stable states of H~. Each of the potential
curves U^(Rh) of Fig. 23.5(a) supports a series of i^(i?h) vibrational
solutions. In Fig. 23.5(b) we show the energies of the lowest lying potential
wells of Fig. 23.5(a). These are the states which are accessible from the
ground state of H~, and the energies agree well with the experimental
observations. In arriving at the radial equation of Eq. (23.13) we assumed
d(/)/dRh to be negligible. The neglected terms couple states on different
potential curves, allowingthe autoionization of He or autodetachment ofH~
doublyexcitedstates. In other words the adiabatic approximation gives
excellent energies, but it does not give the decay rates of doubly excited
states.

Theoretical descriptions 473 12.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0
20.0 22.0 Rl/2(a.u.) 12.0 2.0 4.0 6.0 8.0 10.0 12.0 14.0 16.0 18.0 20.0 22.0
Rh/2(a.u.) Fig. 23.5 (a) Adiabatic potential curves for1P� H~ shown
aseffective quantum numbers vs V/^. One should look edgewise along the the
Wannier ridge line, at v^ = 18"1/4VRh, to see the lowest + channels. In (b),
only the lowest + channels within each n manifold are plotted along with the
levelpositions in each potential. The Wannier ridge asan imaginary straight
line through the avoided crossings is clearly evident in thisfigure(from ref.
12).
474 Rydberg atoms Rau has pointed out that the doubly excited states
inwhich both electrons are at the same distance from the He2+ core have a =
jt/4 and are likely to be at the saddle point of the potential of Fig.
23.4.13Furthermore, if the potential V of Eq. (23.7) is expanded in the
vicinity of 612 = n and a = TCIA it can be approximated by its leading term,
i.e.13 y=ZZo(Z)^ ( 2 3 1 4 ) where Z0(Z) = 2V2(Z � 1/4). For this potential
the energies relative to the double ionization limit are given by ~Z2o(Z) 2{n
+ 5l2f V ; an expression appropriate for the six dimensional coulomb
problem. To account for the higher order terms neglected in Eq. (23.14) a
screening constant o and a quantum defect d are introduced into the
expression for the energy. Explicitly,13 2(n + 5/2-(5)2 v ' ' A similar formula
wasgiven by Read,14 whohas, in addition, summarized muchof the work on
the development of Rydberg formulae for the energies of the double Rydberg
states.15 The formula of Rau, given by the expression of Eq. (23.16) gives an
excellent description of the He~ ls(ns)2 states observed by Buckman et al.16
ifo= 1/2V2 and <5 = 1.67. An interesting aspect of this approach is that all
the nodes in the wavefunction are in the Rh direction, whereas in the
hyperspherical approach the ns2 states have no nodes in the Rh direction, but
are the lowest states in Rh potential curves corresponding to different values
of the set of quantum numbers//. The hyperspherical approach has enjoyed
enormous success in the calculations of energies. It does, however, have the
drawback that the physical interpretation of the set of internal juquantum
numbers is not transparent. Had we used the eigenfunctions of A2, it is not
clear that the situation would be much improved since the individual angular
momenta of the two electrons ix and (2 appear as quantum numbers, and the
� description of Eq. (23.1) and the Stark like splittings of Fig. 23.5 suggest
that ix and �2 are not good quantum numbers. To address this issue methods
of classifying states using group theory were explored by Sinanoglu and
Herrick,17 and they made substantial progress in classifying the states.18
One of the more interesting aspects of this approach is that the notion that the
doubly excited states are related to Stark states arises naturally from the
original group theory work of Park19 relating Stark and angular momentum
states in H. Molmer and Taulbjerg20 make the point that some of the states
described by Herrick are simple linear combinations of products of Stark
states. An approach which joins these two disparate points of view is the
molecular orbital approach of Feagin and Briggs,21'22 who treat the three
body coulomb
Theoretical descriptions 475 problem in which two of the particles have the
same charge, Zx = Z2, and mass m1 =ra2,and the third has a different charge
Z3 and massra3.The approach is similar to the hyperspherical approach in
using an adiabatic approximation. The difference lies in the choice of
coordinates. If rx and r2 are the vectors defining the locations of the two
electrons relative to the He2+, as shown in Fig. 23.1, the two vectors R = r 2
- r ! (23.17a) and r = r! + r2 (23.17b) are introduced. R plays a role
comparable to the role of Rh in the hyperspherical approach. The magnitude
R isthe distance between the two like particles 1 and 2, the two electrons in
He or H~, and r is the distance from the center of mass of particles 1 and 2 to
the third particle, He2+ or H+ for He and H". These coordinates are
precisely the coordinates commonly used to describe H2+. The Schroedinger
equation for the three body coulomb system takes the form22 \2JU12 2ju12y3
R rx r2 where ju12 *s the reduced mass of the twolike particles, ju12 =
mxm2l{ml + m2) = 1/2 for He or H~, and ju123 is the reduced mass of the
center of mass of the two like particles and the third particle; ju12 3 = (mx +
^*2)m3/(mi + m 2 + w3) = 1/2 for He or H~. The eigenfunctions of the atom
must be eigenfunctions of the total angular momentum and its projection on a
space fixed z axis. If we ignore the electron spins for simplicity, the total
angular momentum L and its projection M on the space fixed z axis are
conserved. The spatial wavefunctions ^LM(r,R) are related to the body fixed
wavefunctions by the Euler transformation ^M(r,R) = DUfA0)^f^^(r,/(),
(23.19) where ip,0, and 0 = 0are the Euler angles required to rotate the
spacefixedz axis into the direction of the vector R, the body fixed frame, D^K
*Sarigidt o P wavefunction of total angular momentum L and projections M
and K on the space fixed and body fixed (along R) axes.22 DMK describes
the overall rotation of the atom. (piK(r,R) is the molecular orbital having the
projection of its angular momentum on the R direction equal to number K and
internal quantum numbers represented by /. At a fixed separation R, <piK
describes the motion of He+ relative to the center of mass of the two
electrons. fifc(R) describes the effect of separating the two electrons, which
is analogous to the vibrational motion of the nuclei in a diatomic molecule.
Note that the third Euler angle is zero in D^K{xp,6ft)since it coincides with
the azimuthal motion of r about R.

476 Rydberg atoms We can write the Hamiltonian of Eq. (23.18) as -V2 H =
� + h, (23.20) 2/^ where h = - i + ^ ^ +^ ^ +^ ^ (23.21) 2/^12,3 # '"l >2 is the
Hamiltonian for the motion of the third particle, He2+ or H+, in the body
fixed frame. One of the aspects of the molecular orbital approach which is
most interesting is that the eigenfunctions of the operator h are apparently not
too different from the final atomic eigenfunctions. We shall discuss this point
further, but forthe moment we simply note that the eigenfunctions (piK(r,R)
satisfy the eigenvalue equation hcpiK(r,R) = eiK(R)<piK(r,R). (23.22) The
VR term of the Hamiltonian represents the kinetic energy of the relative
motion of the two electrons. It can be broken into radial and angular parts
using (23.23) 2JU12 2ju12 2ju12 where L^ = -/R X VR. If we now ignore the
off diagonal couplings between the cpiK(r,R) states introduced by LR and the
derivatives of 0iK with respect to R, the Schroedinger equation, Eq. (23.15),
can be reduced to the ordinary differential equation *-Ulk(R)-W\fiK(R) = 0
(23.24) where ufk(R) = h L2 2fi12 <PiK) = �iK{R) + [<PiK 2JU12 </>*)'
(23.25) Eq. (23.24) leads to a series ofvibrational functions in the potential
of UiKL(R) described by Eq. (23.25). A useful feature of the molecular
orbital approach is that the eigenvalue equation ofEq. (23.22) can be
separated in confocal elliptic coordinates,23 and, equally important, these
eigenfunctions are apparently somewhat similar to the final atomic
eigenfunctions.22 The coordinates are given by22 X = (n 4-r2)/R (23.26a)
and /* = (*!- r2)IR. (23.26b) The surfaces of constant pi and X are surfaces
of revolution about the interelectron axis. The eigenfunction (piK(r,R) of Eq.
(23.22) has the form

Theoretical descriptions All <PiK(r,R) = <px( A ) 0 � e ' ^ (23.27) and

hasthe quantum numbers K, n^ and nM, the latter two specifying the number
of nodes in the X and // directions. Since these quantum numbers are good
irrespective of R, they allow the connection between the case R =0 and R
=��. For R^>0, analogous to the united atom limit in a diatomic molecule,
the problem reduces to one with spherical symmetry identical to the H atom,
and rix =n � � - 1, the number of radial nodes, andnM = �� |ra|, the
number of polar nodes. These evolutions are easily envisioned by
considering Eqs. (23.26). As the two electrons are moved together the lines
of constant X become circles, and the lines of constant // radial lines. As /?
�> o� ^ and n^ become directly related to the parabolic quantum numbers,
n1 and n2. Explicitly, n^ = nx and n^ =2n2 if nM is even and 2n2+l if n^ is
odd.22 In a real atom R ^ 0so the actual atomic states have some relation to
the Stark states, as mentioned in the discussion of the initial approach of
Cooper et al.2 and as can be seen in Fig. 23.5(a). An attractive feature of the
molecular orbital approach is that the nodal lines of the
atomicwavefunctionsfall onsurfaces ofconstantju andX. Thisfact seemstobe
true whether the wavefunctions are obtained by a diagonalization technique,
a hyperspherical coordinate approach, or the molecular orbital approach. To
illustrate this point we show in Fig. 23.624 a density plot of a wavefunction
obtained in hyperspherical coordinates for R = 80a0 by Sadegphour and
Greene.12 Although the figure isdrawn using the hyperspherical co-ordinates
012 and a, the linesof constant X andju are shown as well, and itis apparent
that nx = 6 and njU =0. As shown by Rost et al.24 in most cases the nodal
lines of the wavefunctionsfall on surfaces of constant// orA, implying that
theju andX motions are almost always separable. However, the// and R
motions are in general found to be coupled.25 In the discussion of both the
hyperspherical method and the molecular orbital method we assumed that the
motion in the Rh or R direction was slow. While in the Born-Oppenheimer
approximation the slowness isgenerally thought to arise from the much higher
massof the nucleii, as pointed out byFeagin and Briggs,22 it is probably the
repulsive nature of the internuclear potential which allows the separability.
In other words, the success of these methods in treating He isdue to the
interelectronic repusion of the l/r12 potential. Rost and Briggs adopted the
opposite approach, ignoring the electron electron interaction to generate the
diabatic potential U(R)26 Specifically they approximated the Hamiltonian
as26 // =- f ^ - ^ - - - - , (23.28) z z r r ignoring the l/r12 electron electron
interaction. They used trial wavefunctions of the form26 (23.29)

478 Rydberg atoms (a) 0.0 0 0 0.L 0 2 O.'.l 0.4 (b) 0.0 0.0 Fig. 23.6 Contour
plot of adiabatic, twoelectron densities taken from ref. 12. Thenodal (solid)
and antinodal (dashed) linesof constant!andju are overdrawn for (a) the
statewith i^nn) = (0,8)and (b)thestate with (nxn^) =(1,6)(from ref.24).
products of one electron wavefunctions with Z equal to avariational
parameter a. The wavefunction of Eq. (23.29) can also be expressed in
molecular orbital fashion as V = ^f(R)<P(r,R), (23.30) K 0 , (23.31) and the
function U(R) is calculated using U(R) = v? 4 |r - R/2| |r + R/2| i.e. by
integrating over all coordinates save R. Since U(R) depends on a, both
through Z in Eq. (23.31), and through 0, a can be varied to find the minimum
energy of any state. Using nodeless Is trial wavefunctions for xpn and t/V tne
Y found that a = 1.815 gives the minimum He Is2 energy and leads to the He
potential curve of Fig. 23.7. It passes approximately through the avoided
crossings of the adiabatic curves analogous to those shown in Fig. 23.5, and
gives a seriesof

Theoretical descriptions 479 0 -1 -3 P'-S -6 0 2 4 6 8 10 12 R (au) Fig. 23.7

Electronic vibrational potentials U(R) for He (full curve, a = 1.8315) and IT
(broken curve,a = 0.828)(from ref. 26). energies corresponding to increasing
vibrational motion ofR. The energies of the wavefunctions with even
numbers of nodes correspond to the He ns2 states and are in excellent
agreement with energies calculated in other ways. It isinteresting that in these
wavefunctions all the nodes are inf(R), asin the approach of Rau,13 but in the
molecular orbital and hyperspherical wavefunctions most often calculated,
f(R) is nodeless and the nodes appear in other directions. However, in either
case the wavefunction has the largest amplitude at large R, so the presence or
absence of an oscillating wavefunction at small R does not have much
bearing on the energy, although it would make a difference in the excitation
probability from aninitially small atomicstate. The situation is approximately
the same as the H nZ states. All are degenerate, all have the same total
number of nodes, and all have wavefunctions most likely to be found at
similar orbital radii. All of the above approaches are based on separating,
approximately, the motion of theelectrons insome coordinate system.
Aradically different approach is the one used by Richter and Wintgen,27 who
have rewritten the Hamiltonian using coordinate systems which are not
separable but lead to equations simple in form. One of the inherent problems
in all the quantum mechanical approaches is the high density of states near a
double ionization limit. As pointed out by Percival28 and Leopold and
Percival,29 classical techniques offer an attractive alternative, and recent
calculations have demonstrated the usefulness of this approach. For example,
Ezra etal.30have studied the doubly excited He states of total angular
momentum L =0. In these states the motion of the electrons is confined to a
plane. The rather surprising result is that the stable orbits correspond to an
"asymmetric stretch" motion of the two electrons as shown in Fig. 23.8. The

480 Rydberg atoms 50 75 100 Fig.23.8 Probability distribution \^Nn(x,y,z)\2

for the intrashellwavefunction N =n = 6 in the x =0 plane corresponding to
the collinear arrangement r12 = r1+r2. The axes have a quadratic scale to
account for the wave propagation in coulombic systems, where nodal
distances increase quadratically. The fundamental orbit ( ) (as) as well as the
symmetricstretch motion ( ) (ss) alongthe Wannier ridge areoverlayed onthe
figure (from ref. 30). motions of the two electrons are not symmetric about
the He2+, i.e. they are not mirror images of each other. In other words the
motion is not along the Wannier ridge but perpendicular to it. This conclusion
is not really different from those drawn from quantum mechanical
calculations asshown byFig. 23.8,in which itis apparent that the asymmetric
stretch classical orbit falls almost precisely on the antinodes of the quantum
mechanical wavefunction. Laser excitation In addition to the synchrotron
experiments on He, laser experiments have been done using alkaline earth
atoms. The attractions of the laser are two. First, it is in principle possible to
obtain better resolution with alaser than with a synchrotron, and, second, it is
a straightforward matter to start from bound states other than the ground state,
affording access to a wider range of doubly excited states. The drawback to
laser excitation isthat it isnot presently possible to produce tunable radiation
at 200 A, asused in the He experiments. In fact, 2000 A, equivalent toa
photon energy of 6 eV, isa more realistic lower wavelength limit.The
wavelength

Laserexcitation 481 restriction of available lasers makes alkaline earth

atoms, with their low double ionization limits, very attractive. In Ba, for
example, the double ionization limit lies only 10 eVabovethefirstlimit
andcanbereached from abound Rydberg state by the absorption of two 2500
A photons, a technically manageable proposition. The typical experimental
approach is to use a thermal beam of alkaline earth atoms, which is first
excited to a bound Rydberg state by one or two pulsed dye lasers. Typically,
the bound Rydberg state isa high n state, although sometimes it is ahigh
�state as well. The bound Rydberg atom is excited to a high lying doubly
excited state in one of two ways. First it can be excited by a two-photon ICE,
for example a Ba 6snd state can be excited to the 9dnd state. The most
straightforward approach is to use a single laser operating at half the
6snd�> 9dnd transition frequency.31 This approach has the substantial
attraction of simplicity, but a significant drawback as well. Absorption of a
single photon leads to photoionization to the 6p�d continua. Since the
virtual intermediate state of the two photon 6snd-j>9dnd transition is far off
resonance, driving it requires a relatively high laser intensity, and the
undesired single photon ionization is not a negligible problem. An alternative
two photon excitation scheme has been used to excite the Ba 6snd��
7sndtransitions.32 Two frequencies are used, the first coincident with an
overlap integral zero in the 6sml�>6pmi transitons. This approach has the
attraction that the first photon produces no single photon ionization, yet the
virtual intermediate state is insignificantly detuned from the real intermediate
state. It hasthe advantage of requiring lowlaserpowers andproducing
littlesingle photon ionization. However it does have two significant
drawbacks. First, it requires one more laser than does a single color, two-
photon excitation, and second, itfixesthe energy of the virtual state at 2.5 eV
above the first ionization limit, which is rather low. The second approach to
exciting high lying doubly excited states is to make a sequence of single
photon ICE transitions, for example from Ba 6snd �� 6pnd ��
6dnd.Eichmann etal.33 have used this method to excite Ba n{gnod states by
the route 6snod -^ 6pnod�� 6dnod �> 6fnod�> n{gnod, using four dye
lasers. All four transitions are strong single photon transitions which are
easily driven by lasers with 100/J pulse energies. The two problems are the
multiple lasers and the fact that each of the several intermediate autoionizing
states decays, diminishing the signal from the final doubly excited state and
adding to the background ion signal. As isby now apparent, in all these
schemes for exciting high lying autoionizing statesthere is a substantial
number of ionsand electrons created havingnothing to do with the final
doubly excited state. As a result, simply detecting ionization is rarely
adequate, although it is when studying relatively low lyingstatessuch as the
Ba 7sndand 6dnd states. Detection schemes which are only sensitive to the
final n${noto state must beused.A relatively straightforward approach is to
detect only the high energy electrons which the n{�{no�o state ejects in
autoionization to the

482 Rydberg atoms ground state of the ion. For example, in the twophoton
excitation of the Ba 6dnd states from the 6snd states such electrons have
energies of �4.6eV, while those from single photon ionization have energies
of at most 2.3 eV. The success of this approach depends upon there being a
favorable branching ratio for the autoionization to the continua of the ground
state of the ion. As we consider doubly excited states converging to
progressively higher energy states of the ion, autoionization ismore likely to
occur to high lying states of the ion, and detecting the high energy electrons
associated with autoionization to the ground state becomes impractical. As a
result, this approach has only been used on relatively low lying doubly
excited states. Asmentioned above, ifweexcite a Bani�ino�o double
Rydberg state, it is likely to autoionize to an excited nxix state of Ba+, which
is energetically nearby. The other two techniques for selectively detecting the
Ba nxt{noto states depend upon this fact. A laser, often one of the ones used
to produce the Ba ni�ino�o doubly excited state, canbe used tophotoionize
the Ba"1"^^state toproduce Ba2+, which is easily separated from the copious
Ba+ signal to provide a clear signature of the production of the
BanfijiJ^Qstate. This approach assumesthat thedoubly excited state
autoionizes and that it does not give up very much energy to the ejected
electron when it does so. The photon energy of the laser photoionizing the
Ba+ must be high enough to ionize the ions resulting from the final doubly
excited state, but low enough not to ionize background ions produced along
the way. When doubly excited states converging to Rydberg states of the ion
are produced, the product of autoionization is often an ionic Rydberg state,
which is easily ionized by microwave ionization34 or pulsed
fieldionization,33'35 although for field ionization to be useful, risetimes of
~1 JUS are required. Since either of these methods is only sensitive to
Rydberg states of the ion, they provide an excellent means of detecting the
high lying doubly excited states. Experimental observations of electron
correlation The strategy of the experimental study has been to start at the
autoionizing states converging to low lying ionic states, which are
understood using an independent electron picture, and proceed to the doubly
excited states converging to higher lying ionic states, all the while looking
for clear manifestations of correlation of the motion of the two electrons.
Most of the experiments to date have been carried out on doubly excited
states converging to isolated low angular momentum states of the ion. In
general the spectra are well characterized as two photon ICE spectra in
which the observed spectrum is characterized by the product of an overlap
integral and the spectral density of the doubly excited autoionizing states. As
an example we show in Fig. 23.9 the Ba 6sl9d^> 9d/i'd spectra observed by
Camus etal.31 As shown in Fig.

Experimental observations of electron correlation 483 (a) "VYYYYYYY A

(b) (d) -'' 49d5/221d 49d5/220d 49d5/219d Fig. 23.9 Observed excitation
spectra 6sl9d�> 9dn'd in neutral Ba: (b) circular, (c) linear polarization of
laser beams and (d) computed spectra. The sharp resonances A, B and C
correspond respectively to6s1/2 �> 9d5/2,6s1/2�> 9d3/2 and 5d�>
lOgionictransitions usedas standard calibrations with the Fabry-Perot fringes
(a) (from ref. 31). 23.9, there are four peaks in the spectra, corresponding to
the ionic Ba+ 6s1/2 �� 9d3/2 and 6s1/2�> 9d5/2 transitions for n = 20
and 21, in agreement with the calculated spectrum. In spite of the fact that the
spectra can be understood as two-photon ICE spectra, implying that there
isno serious departure from the independent electron picture, the quantum
defects and autoionization widths of these nx�xno�o states display an
interesting dependence on nx. As shown by Bloomfield et al.36 the quantum
defects of both the nxsnos and nxsnod states increase linearly with nx, as
shown by Fig. 23.10. If the quantum defect is due mostly to core penetration,
as it must be when the quantum defects are so large, evaluating the nx
dependence of the expection value of the penetration energy for an outer nos
or nod electron yields a quantum defect which increases with nx as shown in
Fig. 23.10.36 Not only the quantum defects of the nxsnos and nxsnod states,
but also the widths change. The scaled widths no3F of the nxsnos and nxsnod
states are plotted in Fig. 23.11.36 If we first consider the scaled widths of
the nxsnod states, we can see that they rise sharply with nxy a dependence
which is not surprising if we assume that most of the autoionization rate
comes from dipole autoionization to a nearby state of the ion. If the outer nod
electron remains outside the inner nxs electron, the autoionization rate F is
given by

484 Rydberg atoms - D - ne RYDBERG O D O STATES 8 g o: : nd

RYDBERG : STATES 5 � 5 g u. " 6 8 - 2.5 6 7 8 9 10 11 12 CORE
PRINCIPAL QUANTUM NUMBER Fig. 23.10 Quantum defects of the Ba
msns and msnd states as a function of the core principal quantum number m
(from ref. 36). 0.175 z o K 0.125 o 0.1 o 0.025 0.05 O - ns RYDBERG
STATES D - nd RYDBERG STATES 8 9 10 11 12 CORE PRINCIPAL
QUANTUM NUMBER Fig.23.11. Core scalingofautoionization ratesof
theBamsns andmsnd states asa function of the principal quantum number of
the core. The ndstates showincreasing autoionization width with increasing
core size, while thensautoionization rates areindependent of the core
principal quantum number (from ref. 36). T - 2 ^ 8 ^ p>|2 \(n0d\l/r*\ef)\2,
(23.32) where the predominant dependence on n{is through the dipole
moment (ftjS^lftjp) which scales approximately as n{2. Asaresult T^n{4 and
the rapid rise ofno3T with nx shown in Fig. 23.11 is not unexpected. Once
the inner electron n{s orbit

Experimental observationsofelectron correlation 485 becomes large enough

that the outer electron's orbit passes through most of it, so that core
penetration contributes significantly to the autoionization rate, increasing the
size of the inner electron's orbit does not affect the autoionization rate,
because making the n{s orbit larger also makes it more dilute. One of the
ways of defining correlation is that the wavefunctions are not well
represented by the designation n^{nj,o but rather as the + and � states of Eq.
(23.1) or someother linear combination of ni�ino�ostates. Excluding the
configuration mixing of series converging to the twofinestructure levels of an
ionic state, several such cases have been observed. In any atom perturbing
levels converging to higher limits which lie below the first limit often satisfy
this criterion. The Ba 5d7d states are excellent examples. There have also
been observations in more highly excited states. In the Ca 4s7s^7d7s
transition Morita and Suzuki37 observed clear structure due tointeraction
with the 5gnt states. Similarly, Jones et al.38 observed structure in the Ba
5d3/25g / = 2 ^ 4f5/27g / = 3 transition due to the degenerate 6d5/2nf and
6d5/2nh states. In all of these cases there is evidently angular momentum
exchange between the two electrons, as in Eq. (23.1); however, in both cases
the Rydberg electron spends much of itstime far from the ionic core where it
is in a purely coulomb potential. In this sense these states are not highly
correlated states. When the series converging to high lying isolated sates of
Ba+ are excited it is possible to observe clear evidence of correlation. A
good example occurs in the work of Camus et al.,34 who observed the two
photon spectra from the bound Ba 6SAZOP states to the n^sn^ and nxdnop
states. In Fig. 23.12 we showthe two photon spectrum from the Ba6s45p state
to the doubly excited stateslyingnear then = 30 states of Ba+. The spectrum of
Fig. 23.12 corresponds to changing the inner electron from 28s to 35s. The
6s45p �� njd45p and 6s45p ��ftjs45ptransitions for nx< 28appear
assimple two photon ICE transitions, which are for all intents and purposes
identical to the spectrum of Fig. 23.9. As shown by Fig. 23.12, between the
27d45p resonance and the 29s45p resonance there are weak, but easily
observed resonances which correspond to 28pno� and 25fno� states. As
nx is increased thesep and fresonances becomemorepronounced, tothepoint
that the 31pno� resonance is almost as large as the 30d45p resonance. At
energies just above the 30pno� and 31pno� states there is asignal which
isclearly nonzero, and by the time the energy of the 33pno�, state is reached
the spectrum is essentially continuous, i.e. doubly excited states are excited
at any wavelength. By the time we have reached the 32d45p state the outer
electron is not in a pure 45p state; however, we have used the labelfor
simplicity.There aretwo important aspects of these observations;first,the
appearance of the n-pnj, andnxinoi resonances, and, second, which is not
apparent inFig. 23.12,the center of gravity of the resonances shifts from the
ionictransition frequency as the inner electron quantum number nx is
increased. Weshallreturn to thispoint shortly. Eichmann etal.33 and Jones and
Gallagher35 have made observations which are similar to those shown in
Fig. 23.12, so they may be regarded as representative.

486 Rydberg atoms I I 4 cm"1 27d 28d 29s � � | 28s 29d 30s JjhAA*^^
Fig. 23.12 Ba two-photon absorption spectra 6s45p-�n�n'�'. Narrow
resonance lines correspond toionic resonances ofBa+ due tocoincidences
between atomic (Ba) and ionic (Ba+) resonances. They are indicated by an
arrow. Note the evolution to a continuous spectrum at high photon energies
(from ref. 34). Camus et al.34 explained their observations by a picture
which has sometimes been called thefrozen planet model. Qualitatively, the
relatively slowly moving outer electron produces aquasi-staticfieldat
theinner electron givenby l/ro2, and this field leads to the Stark effect in Ba+.
Thefieldallows the transitions tothe n-pno( and/iifno� statesandleads to
shifts of the ionic energies. The presence of the n$not and n{fno�
resonances in the spectrum of Fig. 23.12 is quite evident. Camusetal.
compared the shifts to those calculated in a fashion similar to
aBornOppenheimer calculation. With the outer electron frozen in place at ro
they calculated the Ba+ energies, Wi(ro), and wavefunctions. They then
added the energy Wo(ro) to the normal screened coulomb potential seen by
the outer electron. This procedure leads to aphase shift in the outer electron
wavefunction

Experimental observations of electron correlation 487 3_ CO 1 _ 35 40 45 n

Fig. 23.13Shifts of the centers of gravity compared to the ionic parent
linepositionsfor the observed structures for double Rydberg resonances 6s/tp
�>� 26dn'p (�) and 6s�p �> 27sw'p (#) vsft*;n = 39-47and 50.The
average quantum defect is 4.1 for the 6snp series.The full line is the
theoretical curve (from ref. 34). which varies with ro. Using this approach
they were able to compare the shifts from the ionic lines of the centers of
gravity of the observed resonances. The cross section for a two photon ICE
transition is given by o�cA2(y)(vb� B \vt)29 (23.33) where we have
omitted all terms which are either slowly varying or constant, and vb is the
effective quantum number of the outer electron inthe initial bound state.
Camus et al.34 observed that the resonances shown in Fig. 23.12 are too
wide to come from asinglefinalouter electron state. Rather, each resonance
comes from aband of possiblefinalstates,inwhich caseA2 ofEq. (23.33)is
aconstant, and the cross section is determined by only the squared overlap
integral. The overlap integral is centered on theionic transition frequency.
Thusbyfindingthecenter of gravityof the observed resonanceswe
canfindthetransition frequency ofthe Ba+ ion with the outer electron present,
and we can compare the shifts from the bare ion frequencies to the values
calculated using the procedure outlined in the previous paragraph. In Fig.
23.13 we show the excellent agreement between the measured and calculated
shifts. The frozen planet model is simple and physically appealing. In
addition, it is clearly related to the treatment based on + and - states
originally given by Cooper etal.2The most convincing demonstrations of
thelegitimacy of the frozen

488 Rydberg atoms planet model is one which appears, in retrospect,

obvious. Eichmann etal.33 simply replaced the outer electron bya
staticfieldtocheck that the spectrum isthe same. Using six dye lasers they
excited Ba atoms first to the bound 6s78d states and then tofinaln{�{7Sd
states of n{> 30 viathe route 6s78d-* 6p78d-> 6d78d-> 6f78d-� n{({7Sd,
and field ionized the Ba+ ions resulting from autoionization of the final state.
Their spectra are shown in Fig. 23.14. Fig. 23.14(a) shows the spectrum
obtained by scanning the last laser, driving the 6f78d^ n^{7Sdtransition. It
shows strong background peaks at the location of the ionictransitions to the
n{d and n{g states, mostly due to excitation of Baionsproduced alongtheway.
Fig. 23.14(c) is the spectrum obtained when Ba is photoionized by the first
two lasers, instead of being excited to the 6s78d state, and then excited to
produce the nd and ngRydberg states of Ba+. Fig. 23.14(d) is obtained bythe
same procedure as Fig. 23.14(c) except that afieldof 60V/cm is added.
Thisfieldcorresponds to (r)~2 of the 78d electron. Fig. 23.14(b) is composed
of the normalized superposition of the zero field ion spectrum of Fig.
23.14(c) and the 60 V/cm ion spectrum of Fig. 23.14(d). These two
components should correspond to the background ion signal and the doubly
excited state signals in Fig. 23.14(a). The normalization of Figs. 23.14(c)
and (d) to produce Fig. 23.14(b) effectively allows the amount of background
ion signal in the synthetic spectrum of Fig. 23.14(b) to be adjusted. Finally,
Fig. 23.14(e) is the computed Ba+ spectrum in the 60 V/cm field. Asshown,
the 34g state has disappeared, being replaced by the n = 34 Stark manifold
which is more likely to be excited on its red end than its blue end. The good
match between the actual spectrum of Fig. 23.14(a) and the spectrum of Fig.
23.14(b), synthesized from ion spectra, validates the frozen planet model.
Several other aspects of this experiment are different from the other double
Rydberg experiments. First, the effect of the outer electron is quite noticeable
at a much lower value of n{lno than in the experiments of Camus etal34 and
Jones and Gallagher.35 The reason is that the Ba+ ng states are essentially
degenerate with the higher �states so that much smallerfieldsfrom the outer
electron convert the � states to Stark states. The situation is similar to He,
and is a good example of angular correlation between the two electrons.
Eichmann et al. noted that in the calculated spectrum of Fig. 23.14(e) there is
almost no nd character in the Stark manifold of states and that the pronounced
angular correlation shown in Fig. 23.14(a) could not have been observed if
only theftjd78dstates had been excited, in agreement with other results. The
frozen planet model also agrees with classical calculations of Richter and
Wintgen,39 who find a classically stable state in which both electrons are on
the same side of the atom in orbits which exhibit pronounced angular
correlation, even though the orbital radii of the two electrons are very
different. One of the recurring manifestations of correlation between the two
valence electrons is the conversion of � states to Stark states. Eichmann
etal.40 have studied doubly excited Sr states in which both electrons have
�> 2, so that both inner and outer electrons are in approximately hydrogenic,
easilypolarized states

Experimental observations of electron correlation 489 �e "cc CO 35 36 37

38 39 15680 t tt 15730 15780 3 12 hv6(crrr1) Fig. 23.14(a) Six laser
absorption spectrum ofBarc�78dplanetary states, recorded as B^+ signal vs.
the photon energy, hv6, of the laser driving the last transition, from the 6f5/2
state to the ndlSd and nglM states, (b) Synthetic spectrum as in (a), composed
from a normalized superposition of (c) and (d), including artificial
autoionization broadening of (d). (c) Structure of the purely ionic component
in (a), obtained by six laser excitation of ionic Ba+ nd and Ba+ ng states.
(The extra peak between 33g and 36d appears through accidental excitation
and photoionization of a 8p3/2 state.) (d) Structure of the planetary
component in (a), obtained by repeating the spectrum (c) in the presence of
an external field of 60 V/cm. (e) Theoretical bar spectrum equivalent to (d),
calculated around n = 34 (from ref. 33).

490 Rydberg atoms 1005 0 n \v-Jr / \P: / 1 25 V7f = i If f i

t\hkHi�flItMi:\In11InIw\f\rA 26400 26500 photon energy (crrr1) Fig. 23.15
Observed and calculated squared overlap integral Kv5dlv7f)|2 f�r t n e
transition from the 5d20� state of�> 9 to the lini states as a function of the
photon energy of the laser drivingthetransition. Theexperimental spectrum
(noisytrace)was taken athighlaser power and shows some saturation. The
smooth solid trace is the squared overlap integral calculated using as the
final planetary state the wavefunction ^(^7f), calculated using the potential of
Eq. (23.34)whichcontains aninduced dipole.Thepolarizability aofthe Sr+ 7f
state isindependently calculated to be 5 x lO5^3. The dotted curve isthe
squared overlap integral obtained with a final state calculated using a
coulomb potential (a = 0 in Eq. (23.34)). Note the characteristic phase shift
and height asymmetry of the side maxima of the overlap integral calculated
with the induced dipole potential for the final state when compared tothe
overlapintegralcalculatedwithapurecoulombpotential for thefinalstate
(fromref. 40). with small quantum defects. Specifically, they produced bound
Sr5sno�o states of no ~ 20and �o ~ 9 using the Stark switching technique.
They then further excited these states via the route 5sno�o^> 5pnodo->
5dno�o�> nxinolo. The final nxinoto state, which is a nominal description
only, decays to an excited state of Sr+ which is photoionized to produce
Sr2+. Sr2+ is detected as the wavelength of the laser driving the 5drco�o
�> nxinoto transition is scanned. They observed two clear signatures of the
Stark, or dipole, structure of the doubly excited states; first in the quantum
defects and second, in the overlap integrals. They observed several Rydberg
series converging to excited Sr+ states. From the 5dno = 17�o = 9 state they
observed a series with a quantum defect of 0.70(mod 1) converging to the 6g
state of Sr+. While the observation of a series converging to this limit alone
isindicative of correlation, what ismost interesting is the quantum defect. It is
simply impossible for a Sr coulomb state of �= 9 to have a quantum defect
of 0.70. On the other hand if the outer electron isnot in a

References 491 purely coulomb potential, it is possible for it to have such a

quantum defect. Following Watanabe and Greene41 they assume that the
outer electron is in the potential r(j.o)_-l , 4(4+1) | q ^ ^ where �o and d
depend on ro and are equal to the orbital angular momentum of the outer
electron, �o, and the polarizability of the ion, a, as ro^> oo. Using this
potential they calculate a quantum defect of 6.68 which is in excellent,
perhaps fortuitous agreement with the observations. The second
manifestation of the dipole nature of the outer electron states is found in the
overlap integral observed in thefinalICE transition. For transitions between
coulomb states of effective quantum numbers vx and v2 the overlap integral
has a sin (vx � v2)/(v1 - v2) form. However, for the dipole states it does
not, as shown by Fig. 23.15, a recording of the Sr 5dnot^>7fn'o('otransition.
As shown by Fig. 23.15, the observed spectrum, taken at high laser power, is
not symmetric about the ionicSr+ 5d �> 7ftransition. Rather, itis more
intense on the low frequency side. While the observed spectrum clearly
disagrees with the assumption of the outer electron's being in acoulomb
potential, it is in quite good agreement with the overlap integral assuming
that in the final state the outer electron is in the potential of Eq. (23.34).
References 1. R. P. Madden and K. Codling, Phys.Rev. Lett. 10,516 (1963).
2. J. W. Cooper, U. Fano, and F. Prats, Phys.Rev. Lett. 10,518 (1963). 3. M.
Domke, C. Xue, A. Puschmann, T. Mandel, E. Hudson, D. A. Shirley, G.
Kaindl, C. H. Greene, H. R. Sadeghpour, and H. Peterson, Phys.Rev. Lett. 66,
1306 (1991). 4. P. G. Burke and D. D. McVicar, Proc. Phys.Soc.86,989
(1965). 5. P. L. Altick and E. N. Moore, Phys.Rev. Lett. 15, 100 (1965). 6. T.
F. O'Malley and S. Geltman, Phys.Rev. 137, A1344(1965). 7. L. Lipsky and
A. Russek, Phys.Rev. 142, 59 (1966). 8. J. M. Macek, /. Phys. B 1, 831
(1968). 9. U. Fano, Rep. Prog. Phys. 46, 97 (1983). 10. F. T. Smith,
Phys.Rev. 118, 1058 (1960). 11. C. D. Lin, Phys.Rev. A 10, 1986 (1974). 12.
H. R. Sadegphour and C. H. Greene, Phys.Rev. Lett. 65,313 (1990). 13. A.
R. P. Rau, /. Phys. B 16,L699(1983). 14. F. H. Read, J. Phys. B 10,449
(1977). 15. F. H. Read, /. Phys. B 23, 951 (1990). 16. S. J. Buckman, P.
Hammond, F. H. Read, and G. C. King,/. Phys.B 16,4219(1983). 17. O.
Sinanoglu and D. R. Herrick, J. Chem.Phys.62, 886(1975);65, 850(E)
(1976). 18. D. R. Herrick, Adv. Chem.Phys.52, 1 (1988). 19. D. A. Park, Z.
Phys. 159, 155 (1960). 20. K. Molmer and K. Taulbjerg, J. Phys.B 21, 1739
(1988). 21. J. M. Feagin and J. S. Briggs, Phys.Rev. Lett. 57, 984 (1986).

492 Rydberg atoms 22. J. M. Feagin and J. S. Briggs, Phys.Rev. A 37, 4599
(1988). 23. D. R. Bates and R. G. H. Reid, inAdvances in Atomic and
Molecular Physics, Vol. 4, eds. D.R. Bates and J. Estermann (Academic
Press, New York, 1968). 24. J. M. Rost, J. S. Briggs, and J. M. Feagin,
Phys.Rev. Lett. 66, 1642 (1991). 25. J. M. Rost, R. Gersbacher, K. Richter, J.
S. Briggs, and D. Wintgen, /. Phys. B 24,2455 (1991). 26. J. M. Rost and J.
S. Briggs,/. Phys. B 21, L233 (1988). 27. K. Richter and D. Wintgen, /. Phys.
B 24, L565(1991). 28. I. C. Percival, Proc. Roy. Soc. London A 353, 189
(1967). 29. J. G. Leopold and I. C. Percival, /. Phys. B 13, 1037 (1980). 30.
G. S. Ezra, K. Richter, G. Tanner, and D. Wintgen, /. Phys.B 24, L413
(1991). 31. P. Camus, P. Pillet, and J. Boulmer, /. Phys. B 18,L481 (1985).
32. T. F. Gallagher, R. Kachru, N. H. Tran, and H. B. van Linden van den
Heuvell, Phys.Rev. Lett. 51,1753 (1983). 33. U. Eichmann, V. Lange, and W.
Sandner, Phys.Rev. Lett. 64, 274 (1990). 34. P. Camus, T. F. Gallagher, J.-M.
Lecompte, P. Pillet, L. Pruvost, and J. Boulmer, Phys.Rev. Lett.
62,2365(1989). 35. R. R. Jones and T. F. Gallagher, Phys.Rev. A 42, 2655
(1990). 36. L. A. Bloomfield, R. R. Freeman, W. E. Cooke, and J. Bokor,
Phys.Rev. Lett.53,2234 (1984). 37. N. Morita and T. Suzuki, /. Phys.B 21,
L439(1988). 38. R. R. Jones, P. Fu, and T. F. Gallagher, Phys.Rev.A 44,
4260(1991). 39. K. Richter and D. Wintgen, Phys.Rev. Lett. 65, 1965 (1990).
40. U. Eichmann, V. Lange, and W. Sandner, Phys. Rev. Lett. 68,21 (1992).
41. S. Watanabe and C. H. Greene, Phys.Rev. A 22, 158 (1980).

Index ac Stark shift, 55, 172, 180, 323, 386 molecules, 221 ff anticrossing
spectroscopy, 66, 387-91 rare gas atoms, 216-20 associative ionization,
23SM3 theory, 210 atomic units, 12 collisional ionization autoionization, 95,
398 associative ionization, 242-3 final states, 411-13 by atoms rate, 398, 430
associative ionization, 239-43 n dependence of, 408-9 447-8, 484-5 Penning
ionization, 243 � dependence of, 409-10 by ions, 279 average oscillator
strength, 38 ff by molecules avoided crossing, 89, 95 electron attachment,
230-8 D . r , 1 rotational excitation, 227-9 Balmerformula, 1 A A. -� O1
~en -,M Beutler-Fano profile, 138-9,401,432, 440, c o r e PenetraHon 17,
31, 350 382 .^ F ' � ' � ' core-perturber interaction, 196,243-5 U1 1U , ,-,.
core polarization blackbody radiation J- i_ A- j i r ^ ^ * ^ ~,An m ac Stark
shift, 55-7, 64-5 a d l a b 3 ^c m o d e l f o r 1u a n t u m d e f e c t s ' 348
~50' anticrossing spectroscopy, 66 [o l a r i z a b i l i t i 3 5 1 3 7 7 3 8 3
characteristics, 50-3 A- u +� A 1* I A C + , A . . A. , � .,��., non-
adiabatic model tor quantum defects, photoiomzation by, 60-1, 241 366-73
378 superradiance, 66 coulomb functions, 15, 16,18-20, 415-16 transition
rates, 53-5, 57-9, 60, 65-6 ,. .. 1O o i , . . . . normalization, 19-22
suppression and enhancement in cavities, . , , , ,. o o . X_A numerical
calculation, 22-4 _ D , . o radial expectation values, 25 Bohr atom, 3 . , , ^^.^
D . orin curve crossing model, 272 Born approximation, 200 & Born-
Oppenheimer approximation, 472, 477 , 1 417 depletion broadening,
407,410 cnarrne s, *i / detection of Rydberg atoms i (collision), 417
absorption, 7, 399^401 ) ' c o l l i s i ^a l ionization, 228 ;.i-7 ^.o^ field
ionization, 47-8, 105,207-8, 346, 373-5 open 417 423 fluorescence, 45-7,
207, 345-6 charge densities optogalvanic detection, 453 in correlated states
477-8 photoionization, 345,348 in parabolic states, 73 * ^ d i o d 4 5 3
charge exchange, 27 31,279 diamagnetic interaction, 7, 145-6 final state
distributions, 281-5 ,. , . . , . nn\ � , � i A � * A>T r ^ i i i - dipole matrix
elements, 70-1 classical trajectory Monte Carlo calculations, ,.^ , . . . ^ 9 -L -
_/ dipole states, 490 ir � \A i ? dipole-dipole interaction, 294,314
collisional depopulation Doppler free spectroscopy, 254, 341-2 fine
structure changing collisions, 215-16 TA i * � a^ /, � . ,,. � ~�>rt
Doppler tuning, 36 � mixing collisions, 208 ff A A +4. I~>A , J> � n o f f '
�r . t , . - r- , , dressed states, 324 by atoms, 208 ff; effects of electric fields,
212-13; final states, 211-12 by electrons, 30, 285-7 Einstein coefficient, 39
by molecules, 215 electricfields by ions, 271-2; velocity dependence, 271;
collisions in, 212, 291 final states, 271-2 energies m changing collisions,
211-12 H, 74 n changing collisions alkali atoms, 87; calculation by matrix
alkali atoms, 220-1 diagonalization, 89; avoided crossings, ions, 271-2 89,
95; calculation by quantum defect electrons, 286-8 theory, 140 493

494 Index electric fields (continued) microwave multiphoton transitions

ionization rates experimental, 168-72 H, 83-7 theory, 172-8 nonhydrogenic
atoms, 95-101 microwave spectroscopy, 342-8, 373-5, 384-7
photoexcitation in, 120 molecular orbitals, 474-7 electron attachment, 230-8
Monte Carlo calculations, 277, 283-5, 288 electron capture (see charge
exchange) multichannel quantum defect theory (see electron loss, 276-9
quantum defect theory) electron spectroscopy, 405 nonpenetrating orbits fast
beams, 31, 160-1, 387 fast collisions, 243 ff optical excitation, 32-5, 399
Fermi model of pressure shift, 252 in electric fields, 120 field ionization in
magnetic fields, 7 adiabatic, 106 ff optical spectroscopy, 341,391, 429, 453
by autoionization, 97 ff optical theorem, 204, 251 by tunneling, 85, 96-7
oscillator strength, 38 ff, 120 classical model, 6, 83 overlap integral, 433,
437-9 diabatic, 111-13 of alkali atoms, 97 ff parabolic coordinates, 71 ofY{
83 normalization of wavefunctions, 81 rates'87-8 95 wavefunctions in, 73,77
ff fine structure' 115, 383 parabolic quantum numbers, 72 alkali, 354, 363
Penning ionization, 243 changing collisions, 215-16 Ph a s e shlft constant,
52, 359, 384 coulomb, 16 H e 39i radial 18, 20, 44, 415-16 Floqu'et
approach, 174 photoionization, 33 forced autoionization, 401, 461^
classically stable orbits in magnetic fields, form factor, 200 1 5 3 f f frozen
core approximation, 17 c r o s s sections, 34, 40 ff frozen planet model, 486-
8 hydrogenic structure in electric fields, 124-5, 127-9, 132-5 g factor, 459-
60 in magnetic fields, 147 ff grand angular momentum, 467 nonhydrogenic
structure in electric fields, Hamiltonian 135-41 pje 2SI photon occupation
number, 51 in electric fields, 89 pressure broadening in magnetic fields, 143
alkali-rare gas atoms, 255-63 two valence electron atom, 366, 395 a l k a h
self broadening, 263-6 He +,- states, 467-9, 487 experimental methods, 254-
5 hyperspherical coordinates, 470-4 Rydberg broadening, 266-7 yv F theory,
250-2 impulse approximation, 204 pressure shift Inglis-Teller limit, 75
alkali-rare gas atoms, 255-9 intracollisional interference, 202 alkali self
shift, 263-6 isolated core excitaton (ICE), 402 ff, 433, 440 ff, Fermi model,
7, 253 481 methods, 254-5 � mixing by magnetic fields, 144 theory, 250-3
� mixing collisions (see collisional production of Rydberg atoms
depopulation) c h a r ge exchange, 31-2 Landau-Zener formula, 111, 145,
165-7 collisional optical excitation, 35-7 Legendre polynomials, 76, 99 ff,
122 electron impact, 28-30 level crossing spectroscopy, 358 �Ptical
excitation, 32-5 line broadening (see pressure broadening) spectroscopy,
355-7 Lu-Fano plot, 455-7 J J ^ m changing collisions, 211-12 alkali, 352
microwave ionization alkaline earth, 375-6 circularly polarized microwaves,
190 ff He, 391-2 H in linearly polarized microwaves, 182 ff quantum defect
surface, 419, 430-1, 442-5, 455 other atoms in linearly polarized
microwaves, quantum defect theory, 415 164-8, 178-82 quasi-Landau
resonances, 149 ff

Index 495 R matrix, 415,425 ff, 445, 448-50 scaled energy spectroscopy,
137-8,157 radial functions (see coulomb functions) scattering length,
211,251 radiative collisions, 314 ff Schroedinger equation theory, 322-31 for
atwo electron atom, 366, 396 radiative lifetimes, 39^0, 48,206 for H, 11 in
perturbed series, 457-8 for H in an electric field, 72 reduction by black body
radiation, 45,48 in a magneticfield,143 radiative recombination, 7
selectivefieldionization (seefieldionization), radio recombination lines, 7
323 Ramsey interference fringes, 301, 330 sideband states, 175 resonant
energy transfer spectral density, 433-5 to molecules, 225,229 spherical
harmonics, 13, 14, 367 to other Rydberg atoms, 290 spin-orbit interaction,
115,143, 383 rotational energy transfer, 195, 222ff spin-spin interaction, 383
ionization by, 227-9 Stark switching, 406-7, 410 resonances in, 225-7 state
changing collisions (see collisional resonant Rydberg-Rydberg energy
transfer, depopulation) 291 ff Stuckelberg oscillations, 330 dipole-dipole
theory, 292 ff superradiance, 66 n scaling of widths and cross sections,
303_6 term energies, 342-3 orientational dependence, 306ff trilevel echoes '
255 ' 257 transform limited collisions, 312-13 vibrational energy transfer,
197, 229-30 velocity dependence, 307 Runge-Lenz vector, 146 wave packet
calculations, 153 Rydberg constant, 1 Wigner 3/ symbol,76 Rydberg formula,
2 WKB approximation, 19,80, 125, 149